Kevin Emerson Collins, The Williamson Revolution in Software's

DRAFT 04/01/16—PLEASE DO NOT CITE OR QUOTE WITHOUT THE AUTHOR’S PERMISSION

The Williamson Revolution in Software’s Structure Kevin Emerson Collins∗

Williamson v. Citrix Online altered the threshold test for determining whether a functional claim limitation that not use the term “means” should be construed using the scope-‐narrowing rules of section 112(f). More specifically, Williamson expands the reach of section 112(f) by replacing a formalistic, strong-‐presumption test with a substantive, sufficient-‐structure test. In most arts, Williamson will likely only have a small impact at the margin. Williamson will simply move the needle a bit on the quantitative question about structure in the threshold test: How much structure in a claim limitation is enough to avoid section 112(f)? A bit more structure will be needed after Williamson than was needed before. In contrast, in the software arts, Williamson will have a revolutionary impact. Williamson will finally force the Federal Circuit to address the definitional question about software’s structure in the threshold test: What constitutes structure in a software limitation in the first place?

This essay makes two basic points. First, it demonstrates that Williamson will trigger a significant change in how section 112(f) applies to software patents. Before Williamson, the Federal Circuit was able to duck the difficult, definitional “What is structure?” question in almost all of its software cases, thanks to a combination of the formalistic, strong-‐presumption threshold test and a purely relative definition of structure in the rule for identifying corresponding structure in the specification. Now, after Williamson, the Federal Circuit must either acknowledge that the functional limitations of a software claim have no structure and that all functional software limitations are governed by section 112(f), or it must develop a new, stand-‐alone definition of software’s structure. Second, this essay maps the difficult road that lies ahead if the Federal Circuit chooses to develop a stand-‐alone definition of software’s structure.

∗ Professor of Law, Washington University.


1 Collins – Williamson – Draft 04/01/16

The Williamson Revolution in Software’s Structure Kevin Emerson Collins∗

I. History, Policy, and Doctrine ................................................................................................................ 9

II. The Section 112(f) Threshold Test ..................................................................................................... 14

A. Intent Theory Versus Scope Theory ............................................................................................... 15

B. Uncertainties in the Quantitative Question after Williamson ....................................................... 20

III. Section 112(f) and Software .......................................................................................................... 24

A. The Software/Section 112(f) Mismatch ......................................................................................... 25

B. Algorithms as Corresponding Structure ......................................................................................... 30

IV. The Definitional Question after Williamson: Pre-‐Williamson Doctrine Is Unhelpful ..................... 36

A. Algorithms as Corresponding Structure as a Model? .................................................................... 38

1. The Federal Circuit’s Relative Definition of an Algorithm .......................................................... 40

2. A Relative Definition of Structure in the Threshold Test ........................................................... 43

B. The Pre-‐Williamson Threshold Test as a Model? ........................................................................... 46

1. Circuit as Physical Structure ....................................................................................................... 47

2. Inputs and Outputs as a Clue to Logical Structure ..................................................................... 50

V. Towards a Stand-‐Alone Definition of Software’s Structure ............................................................... 55

Conclusion ................................................................................................................................................. 62

In the first half of the twentieth century, the Supreme Court declared a number of

patent claims using purely functional limitations—that is, limitations not reciting any of the

structural properties of the technology that an inventor actually invented—to be facially invalid

for overbreadth.1 In response, Congress adopted what is now section 112(f) as part of the 1952

∗ Professor of Law, Washington University. 1 See infra __.



Patent Act.2 Section 112(f) expressly authorizes patent drafters to draft claims with purely

functional limitations: a claim limitation “may be expressed as a means … for performing a

specified function without the recital of structure [or] material … in support thereof.”3

However, as a price for using purely functional language, it mandates that the authorized

limitations be construed using a statutorily specified, scope-‐narrowing rule of claim

construction. Section 112(f) limitations “shall be construed to cover [only] the corresponding

structure [or] material … described in the specification and equivalents thereof.”4

In Williamson v. Citrix Online, the Federal Circuit altered the threshold test for

determining whether a claim limitation is subject to section 112(f).5 In the decade preceding

Williamson, the Federal Circuit developed formalistic, strong presumptions that usually allowed

a patent drafter to opt in or out of section 112(f). If a claim limitation contained the words

“means for” performing a function, section 112(f) almost always governed, but, if it employed a

synonym for “means,” such as a “device for” or “mechanism for,” performing the function,

section 112(f) almost always never governed.6 Williamson expanded the reach of section 112(f)

by reducing the strength of the latter presumption that limitations without the term “means

for” are not section 112(f) limitations. That presumption can now be overcome with the

2 35 U.S.C. § 112(f) (2012). This Essay refers to the statute as section 112(f), the title given to the statute by the AIA. Prior to the AIA, the statute was known for many years as section 112, paragraph 6. 3 35 U.S.C. § 112(f) (2012). 4 Id. 5 Williamson v. Citrix Online, LLC, 792 F.3d 1339 (Fed. Cir. 2015) (en banc in part). 6 Id. at 1348–49.



substantive showing that a limitation recites “function without reciting sufficient structure for

performing that function.”7

In the mechanical arts, Williamson will likely only lead to incremental change.

Williamson will move the needle on the quantitative question about structure in the section

112(f) threshold test. How much structure is enough to avoid the scope-‐narrowing rules of

section 112(f)? After Williamson, a bit more structure is needed than was needed before.

However, with respect to software patents in particular, Williamson mandates revolutionary

change. To bring Williamson to bear on software patents, the Federal Circuit will be forced to

openly ask and answer a more fundamental question that, to date, is has ducked. This more

fundamental question is the definitional question about software’s structure in the threshold

test: What properties of a software program should count as structure to begin with when they

are recited as claim limitations? The goals of this essay are to highlight the sui generis nature of

the definitional question in the software arts; prove that, when posed as part of the threshold

test, the definitional question raises a novel doctrinal issue that the Federal Circuit did not

squarely confront in its pre-‐Williamson software cases;8 and map out the difficult path forward

that the Federal Circuit must follow in the post-‐Williamson era to provide a meaningful answer

to the definitional question of the threshold test in software.

The definitional question is clearly important as a conceptual, logical matter. Knowing

what constitutes structure in a claim limitation is a precondition for tallying the amount of

structure recited and assessing whether that amount is sufficient to avoid section 112(f). What

7 Id. at 1349 (citing Watts v. XL Sys., Inc., 232 F.3d 877, 880 (Fed. Cir. 2000)). 8 There are exceptions. See infra Section IV.B.



might be less clear, however, is why the definitional question doesn’t have a simple answer in

software. In mechanical technologies, it does have a simple answer. What constitutes structure

is usually intuitive to a technological neophyte, let alone an experienced patent judge: structure

includes the physical, spatial, and material properties of a technology. Here, the answer to the

quantitative “How much structure is enough?” question may involve intricate legal reasoning,

but the answer to the definitional “What is structure?” question does not. Software, however,

is an intrinsically odd type of technology. The physical, structural qualities of a software

invention are either the open or closed gates of a programmed computer or the arrangement

of electrons on a storage medium. Yet, these structural qualities are irrelevant to the definition

of what a software inventor has actually invented. Software has been engineered with the

express goal of allowing programmers to remain ignorant of the physical structure of their

inventions, meaning that software inventions can only reasonably be defined by reciting what

the software does or how it performs in functional terms. There is no relevant, physical

structure on which an economically rational patent regime can rely to curtail the scope of

functional claims. To bring section 112(f) to bear on software, the Federal Circuit must develop

a sui generis definition of logical structure in software—a definition which, importantly, invokes

metaphorical structure in the sense that it is still a functional description of how software

performs, not literal structure in the sense of the physical properties of a programmed

computer or software recorded on a storage medium.

To be clear, the ideas that software lacks relevant physical structure and that section

112(f) requires a sui generis definition of software’s structure had been recognized well before

Williamson. Ever since its 1999 opinion in WMS Gaming, Inc. v. International Game



Technologies, the Federal Circuit has held that, in the context of searching for the

corresponding structure in the specification, software’s structure is something akin to logical,

rather than physical, structure: it is an “algorithm” or a step-‐by-‐step procedure specifying how

to perform the function recited as a claim limitation.9 However, what has not been recognized

is that the Federal Circuit’s algorithm jurisprudence is of little to no use for answering the

definitional question in the context the section 112(f) threshold test. Although the Federal

Circuit’s case law on algorithms might lead even an informed reader to presume that the

definitional question about software’s structure has already been fully asked and answered

prior to Williamson, it has in fact been answered in a manner that is specific to identifying

corresponding structure in the specification and that does not give the Federal Circuit any

traction when identifying structure in claim limitations.

The underlying reason for this lack of traction is that the Federal Circuit has to date only

offered a relative definition of an algorithm, and thus of software’s structure, not a stand-‐alone

definition. Again, an algorithm is step-‐by-‐step procedure specifying how to perform a function

recited as a claim limitation. Whether a specification recites an algorithm can only be

determined by comparing the series of functions disclosed in the specification to the baseline

functionality recited as a claim limitation. A reader cannot examine a series of functions in

isolation and say that it is or is not an algorithm. Corresponding structure is defined only in

relation to a particular claim; the exact same disclosure may be structure in relation to a more

general claim but not in relation to a more specific claim. During the search for corresponding

structure, this relative definition of software’s structure may be a bit awkward at times, but it at

9 WMS Gaming Inc. v. Int’l Game Tech., 184 F.3d 1339, 1347–50 (Fed. Cir. 1999).



least provides workable rule. However, during the search for structure in the claims, it is

conceptually bankrupt. There is no self-‐evident baseline to which to compare the functional

claim language. If an algorithm is what counts as structure in a software claim, for what task,

precisely, must functional claim limitations be a step-‐by-‐step procedure?

Before Williamson, the strong presumptions allowed the Federal Circuit to sweep the

difficult task of answering the definitional “What is structure?” question in the section 112(f)

threshold test under the rug. They allowed the Federal Circuit to suppress the true extent of the

sui generis nature of the law needed to bring section 112(f) to bear on software. They created

the appearance that the application of section 112(f) to software claims was at least close to

situation normal, when it is in fact anything but. After Williamson, however, the curtain will be

pulled back, and the Federal Circuit will be forced to confront the definitional question about

software’s structure in the threshold test head-‐on. This confrontation will lead to a revolution

in the doctrine of section 112(f) as applied to software, although the nature of that revolution

remains to be seen.

A first, and doctrinally simple, revolution would be to fully embrace a relative definition

of structure that results from defining software’s structure as an algorithm. Perhaps patent

drafters can disclose structure in the specification of a software patent in the form of an

algorithm, but perhaps they cannot recite structure in the limitations of a software claim.

Following this route, Williamson would subject every functional software limitation to the

scope-‐narrowing rules of section 112(f), regardless of the level of specificity at which the

limitation is drafted, and vastly expand the reach of section 112(f).



A second, and doctrinally more difficult, revolution would be to formulate the stand-‐

alone definition of software’s structure that is needed if there is ever to be any structure in

functional software claims. Perhaps the Federal Circuit will finally identify a level of specificity

or granularity that is required for a functional description of a software invention to count as

logical structure, regardless of whether the functional description occurs in the specification or

the claims, and, if it occurs in the specification, regardless of the generality of the claims in need

of corresponding structure. The formulation of a stand-‐alone definition of structure in software

for the purpose of the section 112(f) threshold test would be a difficult undertaking.10 The

echoes of Learned Hand’s levels-‐of-‐generality test for drawing the idea/expression dichotomy

in copyright are clear, and that test is notorious for its lack of ex-‐ante clarity.11 A stand-‐alone

definition of structure is what Mark Lemley has alluded to with the distinctions in software

patents between a “goal” and “steps for achieving a goal,”12 a “goal” and a “way of

implementing a goal,”13 and the “problem” the patentee solved and “what the patentee …

actually did.”14 It could be based on what I have elsewhere described as a line between

functional limitations reciting end-‐user preferences that should be subject to section 112(f) and

functional limitations reciting how those end-‐user preferences are achieved that should not.15

Regardless of its verbal formulation, the initial step for developing the needed doctrine should 10 It is precisely the difficulty of formulating such a stand-‐alone definition of software that has led the Federal Circuit to give only lip service to the statutory mandate in section 112(f) for step-‐plus-‐function method claims. See infra __. 11 Nichols v. Universal Pictures Corp., 45 F.2d 119 (2d Cir. 1930); see infra __. 12 Mark A. Lemley, Software Patents and the Return of Functional Claiming, 2013 Wis. L. Rev. 905, 907. 13 Id. at 947 14 Id. at 963. 15 Kevin Emerson Collins, Patent Law’s Functionality Malfunction and the Problem of Overbroad, Functional Software Patents, 90 Wash U. L. Rev. 1399, 1421–23, 1466 (2013).



be an interdisciplinary conversation that brings the expertise of lawyers, computer scientists,

and economists to bear on the problem in order to sketch the possible levels of abstraction at

which a functional description of software could be deemed to be logical structure for the

purpose of section 112(f).16

This essay proceeds as follows. The first three parts provide background. Part I briefly

summarizes the history, policy, and doctrine section 112(f). Part II discusses the section 112(f)

threshold test in greater detail and explains how Williamson altered that test. Turning to

software patents before Williamson, Part III identifies why the application of section 112(f) to

software requires sui generis rules that focus on logical structure, and it reviews the Federal

Circuit’s cases that require the disclosure of an algorithm in the specification as the

corresponding structure for section 112(f) software limitations. The final two parts address life

after Williamson. Part IV argues that the Federal Circuit’s pre-‐Williamson software cases

provide little guidance for courts seeking to answer the definitional “What is structure?”

question as part of the threshold test and that Williamson therefore mandates revolutionary

change in section 112(f) as it applies to software. Either practically all software limitations

become section 112(f) limitations, or the Federal Circuit must develop a stand-‐alone definition

of structure. Part V briefly discusses the difficulty of the latter task of defining the structure of a

software invention in a free-‐standing manner.

16 Id. at 1466–67.



I. History, Policy, and Doctrine

During the first half of the twentieth century, the Supreme Court invalidated a number

of claims that employed purely functional claim language to identify the outer bounds of an

inventor’s rights. In these cases, the Court repeatedly stated that functional claims were

problematic as a policy matter because they granted patentees rights that were overbroad with

respect to the patentees’ actual contributions to progress.17

In 1928, Holland Furniture Co. v. Perkins Glue Co. involved a patent on a new starch-‐

based glue with a low water content that had previously been attainable only in glue made

from animal fat.18 The inventor obtained a broad claim to any starch-‐based glue “having

substantially the properties of animal glue.”19 The Court noted that the claim described the

scope of the patentee’s rights “not in terms of [the product’s] own physical characteristics or

chemical properties . . . but wholly in terms of the manner of use of the product,”20 and it

invalidated the claim because the functional language lead to overbreadth.21 “As a description

of the invention, [the functional language] is insufficient, and, if allowed, would extend the

monopoly beyond the invention.”22

17 Due to the rhetoric in which the opinions are cast, these cases are sometimes mistakenly labeled as indefiniteness, rather than overbreadth, cases. Collins, supra note __, at 1429 n.122. 18 277 U.S. 245, 247 (1928). 19 Id. at 250. 20 Id. at 256 21 Id. at 257 (“A claim so broad, if allowed, would operate to enable the inventor, who has discovered that a defined type of starch answers the required purpose, to exclude others from all other types of starch, and so foreclose efforts to discover other and better types. The patent monopoly would thus be extended beyond the discovery, and would discourage rather than promote invention.”). 22 Id. at 258.



A decade later, General Electric Co. v. Wabash Appliance Corp. again invalidated a

functional claim for overbreadth.23 A General Electric employee produced a new tungsten

filament for a light bulb that had two commercially valuable properties—it suffered from

neither sagging nor offsetting24—and General Electric obtained a functionally defined patent

claim that encompassed tungsten filaments “made up mainly of a number of comparatively

large grains of such size and contour as to prevent substantial sagging and offsetting during a

normal or commercially useful life for such a lamp or device.”25 The Court held the claim

“invalid on its face” because of its functional language granted the inventor overbroad rights.26

“The claim . . . falls within the condemnation of the doctrine that a patentee may not broaden

his product claims by describing the product in terms of function” and “vividly illustrates the

vice of a description in terms of function.”27

Finally, in 1946, the Supreme Court issued the most famous of its functional-‐language

cases, Halliburton Oil Well Cementing Co. v. Walker.28 Mr. Walker made an improved apparatus

for measuring the depth of oil wells. Prior art devices measured the time at which sound waves

sent into the well returned to the top of the well as echoes,29 and Mr. Walker added a

“mechanical acoustical resonator” to the echo-‐measurement device that made it easier to

23 304 U.S. 364 (1938). 24 Id. at 366–367. 25 Id. at 368. 26 Id. 27 Id. at 371. More specifically, the Court prohibited the use of “conveniently functional language at the exact point of novelty.” Id.; see also Halliburton Oil Well Cementing Co. v. Walker, 329 U.S. 1, 8 (1946). 28 329 U.S. 1 (1946). 29 Id. at 4–5.



record the faint echoes from the small protrusions into the well created by the tubing collars or

couplings at pipe junctions.30 The functional claims before the Supreme Court recited “means

associated with [the prior-‐art] device for tuning said receiving means to the frequency of

echoes from the tubing collars of said tubing section to clearly distinguish the echoes of said

couplings from each other.”31 Focusing on the “overhanging threat of the functional claim,” the

Court invalidated it:32

[W]hat [Walker] claims here is that his patent bars anyone from using in an oil well any device heretofore or hereafter invented which combined with the [prior art] performs the function of clearly and distinctly catching and recording echoes from tubing joints with regularity. Just how many different devices there are of various kinds and characters which would serve to emphasize these echoes, we do not know... In this age of technological development there may be many other devices beyond our present information or indeed our imagination which will perform that function and yet fit these claims.33

These three Supreme Court cases barring purely functional claims because of

overbreadth remain good law today in the sense that a patent applicant who is the first to

invent a technology that performs a function cannot claim all devices that can perform that

function. However, Congress softened the impact of the Court’s functional-‐claiming cases by

enacting what is now section 112(f) as part of the 1952 Patent Act:

An element in a claim . . . may be expressed as a means or step for performing a specified function without the recital of structure, material, or acts in support thereof, and such claim shall be construed to cover the corresponding structure,

30 Id. at 6–7. 31 Id. at 8–9. 32 Id. at 12. 33 Id. at 12.



material, or acts described in the specification and equivalents thereof.34

Section 112(f) allows patent drafters to use functional claim limitations, which can be helpful to

ensure that patents play their expected role of providing an incentive to innovate because

describing an invention without using functional limitations is sometimes difficult and

invalidating any claim with purely functional limitation might leave inventors without effective

patent protection. However, section 112(f) exacts a price from patent drafters who use

functional limitations. It codifies an exception to the default rule of claim construction: the

permitted functional claim language refers only to devices that have the “corresponding

structure” or “material” that the patentee discloses in the specification, as well as its

equivalents.35 A limitation under section 112(f) is usually significantly narrower than the same

limitation would be if it were to be construed under the default rules of claim construction

specified in Phillips v. AWH because, under section 112(f), claim scope is hewn much more

closely to the particular embodiments disclosed in the specification.36

Today, the process of claim construction under section 112(f) proceeds in three steps.

The first step is a threshold test: Is any given limitation the type of limitation that is subject to

34 35 U.S.C. § 112(f) (2012). 35 Id. The Supreme Court’s cases focused on the vice of functional claiming at the point of novelty. See supra note __. However, Congress’s response to these cases was not limited to functional limitations at the point of novelty. Section 112(f) on its face applies to all functional limitations. There are good reasons to believe breadth at a claim’s point of novelty is far more problematic than breadth at limitations aspects of prior art technologies. Kevin Emerson Collins, Getting into the “Spirit” of Innovative Things: Looking to Complementary and Substitute Properties to Shape Patent Protection for Improvements, 26 Berkeley Tech. L.J. 1217 (2011); Lemley, supra note __, at 958–59. This essay, however, brackets the issue of whether section 112(f) should have more bite at a claim’s point of novelty and leaves it for another day. 36 Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc). Patent owners also often prefer having Phillips rather than section 112(f) govern the construction of a claim limitation because section 112(f) limitations often lead to claims being held invalid for indefiniteness. See infra __.



section 112(f), or should Phillips determine the meaning of the claim language?37 The second

step employs the default rules of claim construction to interpret the meaning of the functional

language employed in the claim.38 The third step then identifies the “corresponding structure”

disclosed in the specification that is capable of performing the disclosed function.39 In order to

infringe, a technology must literally perform the function specified in the claim limitation, and

its structure must be that of the corresponding structure disclosed in the specification or its

equivalents.

If the specification does not disclose any corresponding structure, then the claim is

invalid for indefiniteness under section 112(b) of the Patent Act.40 The indefiniteness doctrine

holds that claims that employ limitations whose meanings cannot be ascertained with

reasonable certainty are invalid.41 Indefiniteness is a common-‐sense rule: the scope of a claim

to a “thingamajig” cannot be ascertained, so there is poor public notice, and there are many

instances in which neither the validity nor infringement analyses can proceed. Section 112(f)

states that the meaning of a functional claim limitation is the corresponding structure in the

37 The threshold test is addressed at greater length below. See infra Section II. 38 Generation II Orthotics, Inc. v. Med. Tech., Inc., 263 F.3d 1356, 1364–65 (Fed. Cir. 2001). 39 That is, the specification must contain descriptive text by which a person of skill in the field of the invention would “know and understand what structure corresponds to the means limitation.” Finisar Corp. v. DirecTV Grp., Inc., 523 F.3d 1323, 1340 (Fed. Cir. 2008). An enabling specification is not enough. Biomedino LLC v. Waters Techs. Corp., 490 F.3d 946, 952 (Fed.Cir.2007) (“The inquiry is whether one of skill in the art would understand the specification itself to disclose a structure, not simply whether that person would be capable of implementing a structure. Accordingly, a bare statement that known techniques or methods can be used does not disclose structure.”) (citations omitted). The disclosed structure must also be clearly linked to the function recited in the claim limitation. B. Braun Medical, Inc. v. Abbott Laboratories, 124 F.3d 1419, 1424 (Fed. Cir. 1997). 40 In re Donaldson Co., 16 F.3d 1189, 1195 (Fed. Cir. 1994) (en banc). 41 Nautilus, Inc. v. Biosig Instruments, Inc., 134 S. Ct. 2120 (2014). Indefiniteness derives from the statutory requirement that a patent “particularly point[] out and distinctly claim[] the subject matter which the inventor regards as the invention.” 35 U.S.C. § 112(b) (2012).



disclosure and its equivalents, so the limitation has no discernable meaning if there is no

corresponding structure.42

By its literal terms, section 112(f) also governs method claims. That is, a claim limitation

“may be expressed as a … step for performing a specified function without the recital of … acts

in support thereof, and such [method] claim shall be construed to cover the corresponding …

acts described in the specification and equivalents thereof.”43 While the Federal Circuit has

occasionally discussed the notion of a step-‐plus-‐function method claim that is subject to section

112(f), it has not yet met a method claim that it believed should be subject to section 112(f).44

For this reason, section 112(f) is often equated exclusively with “means-‐plus-‐function”

limitations in product or system claims.

II. The Section 112(f) Threshold Test

The section 112(f) threshold test determines whether a claim limitation is subject to the

default rule of claim construction articulated in Phillips or the scope-‐narrowing rule of claim

construction specified in section 112(f). Section II.A discusses the intent and scope theories of

the threshold test that have jockeyed for importance over the last two decades, and it reports

how Williamson v. Citrix Online elevates scope theory over intent theory. Looking at the

mechanical arts, Section II.B discusses several difficulties that Williamson will create for the

42 Donaldson, 16 F.3d at 1195. 43 35 U.S.C. § 112(f) (2012). 44 Seal-‐Flex, Inc. v. Athletic Track and Court Const., 172 F.3d 836, 848–51 (Fed. Cir. 1999) (Rader, J, concurring); O.I. Corp. v. Tekmar Co. Inc., 115 F.3d 1576, 1582–84 (Fed. Cir. 1997).



Federal Circuit when it seeks to answer the quantitative “How much structure is enough?”

question wound up with the threshold test.

A. Intent Theory Versus Scope Theory

Two competing theories of the reach of section 112(f) have struggled for dominance

over the last several decades. Intent theory suggests that section 112(f) is an option that

Congress placed at the disposal of patent drafters and that section 112(f) should apply when,

and only when, patent drafters use the word “means” as a signal that they intend to invoke it.

In contrast, scope theory suggests that the section 112(f) threshold test should look beyond the

patent drafters’ semantic word choice and focus on whether the claim limitation recites enough

structure so that the Supreme Court’s policy concerns about overbroad claim scope raised in

Holland Furniture, General Electric, and Halliburton Oil Well Cementing are mitigated or

eliminated. Neither intent theory nor scope theory has ever achieved complete dominance in

the Federal Circuit, but the weight given to each has varied over time. In the decade or so prior

to Williamson, intent theory had the upper hand. With Williamson, however, the Federal Circuit

established the dominance of scope theory.

The choice between the intent and scope theories reflects many of the policy tradeoffs

wound up with the classic choice between rules and standards.45 Like other rules, a threshold

test based on intent theory produces greater certainty. It “provides clarity and predictability for

the public and the patentee alike. It helps the public to determine when claim elements are

expressly limited to structures disclosed in the specification (or their equivalents) and provides 45 Frederick Schauer, Playing by the Rules: A Philosophical Examination of Rule-‐Based Decision-‐Making in Law and in Life (1991).



the patentee with the tools for reliability invoking or avoiding means-‐plus-‐function claiming.”46

However, also like other rules, it is under-‐inclusive with respect to the justifiable outcome

because it frequently sanctions the very functional claims that the Supreme Court viewed as

economically overbroad.47 Inversely, scope theory yields a more difficult-‐to-‐administer

standard that imposes a restriction on patentability that is more closely tailored to eliminating

the overbreadth of functional claims.

In 1996, both intent and scope theory were given first billing in different opinions by

different panels. In Greenberg v. Ethicon Endo-‐Surgery, the Federal Circuit articulated a strong

version of intent theory in the course of deciding that the term “detent mechanism” in a

medical-‐instrument claim was not subject to section 112(f).48 Drawing on the literal text of

section 112(f), the court argued that section 112(f):

provides that an element in a claim for a combination ‘may be expressed’ as a means for performing a function, which indicates that the patentee is afforded the option of using the means-‐plus-‐function format. The question then is whether, in the selection of claim language, the patentee must be taken to have exercised that option.49

Thus, intent theory gives rise to twin presumptions of threshold test: “the term ‘means’

(particularly as used in the phrase ‘means for’) generally invokes section [112(f)] and … the use 46 Apple Inc. v. Motorola Inc., 757 F.3d 1286, 1297–98 (Fed. Cir. 2014). 47 See supra __. The explanation of why functional claims are overbroad is more complicated that is commonly acknowledged. Collins, supra __, at 1411–24. 48 Greenberg v. Ethicon Endo-‐Surgery, Inc., 91 F.3d 1580 (Fed. Cir. 1996). 49 Id. at 1584; see also York Prods., Inc. v. Central Tractor Farm & Family Ctr., 99 F.3d 1568, 1574 (Fed. Cir. 1996) (“In determining whether to apply the statutory procedures of section 112, ¶ 6, the use of the word ‘means’ triggers a presumption that the inventor used this term advisedly to invoke the statutory mandates for means-‐plus-‐function clauses.”); cf. Donald Chisum, Chisum on Patents § 18.03[5][e][ii][E] (“To apply Section 112/6’s claim construction to a claim phrase that does not use the key word means is, potentially, a significant step beyond the strict, literal language of the statute.”).



of a different formulation generally does not.”50 In contrast, in Cole v. Kimberly-‐Clark Corp., the

majority opinion articulated a version of scope theory to explain the policy behind the section

112(f) threshold test. The court downplayed intent theory. “Merely because a named element

of a patent claim is followed by the word ‘means,’ … does not automatically make that element

a ‘means-‐plus-‐function’ element,” and “merely because an element does not include the word

‘means’ does not automatically prevent that element from being construed as a means-‐plus-‐

function element.”51 Instead, the court stated that claim limitations should be subjected to the

scope-‐narrowing rules of section 112(f) threshold test if the term at issue is excessively

functional and thus overbroad. That is, a section 112(f) claim limitation “must not recite a

definite structure which performs the described function.”52

By the 2000s, intent theory had clearly gained the upper hand. In Lighting World, Inc. v.

Birchwood Lighting, Inc., “the presumption flowing from the absence of the term ‘means’” had

become “a strong one that is not readily overcome.”53 In fact, the articulation of intent theory

in Lighting World suggested that scope theory was entirely defunct because a claim limitation

not using the word “means” might not be a section 112(f) limitation even if it encompassed all

50 Greenberg, 91 F.3d at 1584; see also 1583 (noting that, if claims use the term “means for,” the Federal Circuit noted that the drafters “intend to invoke” section 112(f) “and that there is therefore seldom any confusion about whether [112(f)] applies to a particular element.”). 51 Cole v. Kimberly-‐Clark Corp., 102 F.3d 524, 531 (Fed. Cir. 1996). 52 Id. Judge Rader, who had previously authored an opinion emphasizing intent theory in York Prods. v. Central Tractor Farm & Family Ctr., see supra note __, dissented in Cole, arguing that intent theory should animate the section 112(f) threshold test. [] Scope theory was arguably quite novel at this time. In its 1998 opinion in Mas-‐Hamilton Group v. LaGard, Inc., 156 F.3d 1206, 1213–14 (Fed. Cir. 1998), the Federal Circuit held for the first time that a phrase not using “means” was governed by section 112(f). [Chisum]; cf. Lighting World, Inc. v. Birchwood Lighting, Inc., 382 F.3d 1354, 1359–60 (Fed. Cir. 2004) (labeled Mas-‐Hamilton as an “exceptional case”). 53 Lighting World, Inc. v. Birchwood Lighting, Inc., 382 F.3d 1354, 1358 (Fed. Cir. 2004).



structures that are capable of performing the claimed function.54 The strength of intent theory

was subsequently reinforced with the assertion that “[w]hen the claim drafter has not signaled

his intent to invoke § 112 ¶ 6 by using the term ‘means,’ we are unwilling to apply that

provision without a showing that the limitation essentially is devoid of anything that can be

construed as structure.”55 Even under the strong-‐presumption formulation of the threshold

test, the Federal Circuit’s opinions are laced with scope-‐theory sound bites.56 However, very

few claim limitations that did not employ the word “means” were held to actually be governed

by section 112(f).57

In 2014, the conflict between intent and scope theories again resurfaced in the

disagreement between the Federal Circuit’s majority and dissenting opinions in Apple v.

Motorola.58 The majority adopted a strong version of intent theory to reverse a district court

holding that a claim limitation without the term “means” was not a means-‐plus-‐function claim.

“The strong presumption created by not including means in a claim limitation … signals to the

court that the patentee has chosen to … avoid[] the benefits of Section 112, ¶ 6.”59 In contrast,

54 Id. at 1361–62 (noting that the term “connector” is “certainly broad” and that it “may in the end include any structure that performs the role of connecting”). 55 Flo Healthcare Solutions, LLC v. Kappos, 697 F.3d 1367, 1374 (Fed. Cir. 2012) 56 Lighting World, 382 F.3d at 1360 (noting that section 112(f) “does not apply to a term that is simply a nonce word or a verbal construct that is not recognized as the name of structure and is simply a substitute for the term ‘means for’”). 57 But see Welker Bearing Co. v. PHD, Inc., 550 F.3d 1090, 1096–97 (Fed. Cir. 2008) (“mechanism”); MIT v. Abacus Software, 462 F.3d 1344, 1355 (Fed. Cir. 2006) (“mechanism”); Mas-‐Hamilton Group v. LaGard, Inc., 156 F.3d 1206, 1213–14 (Fed. Cir. 1998) (“element” and “member”). 58 Apple Inc. v. Motorola, Inc., 757 F.3d 1286, 1297 (Fed. Cir. 2014). 59 Id. at 1297. The majority, however, seems to have forgotten about the Supreme Court’s functional-‐language cases from the first half of the twentieth century in its explanation of the consequences that a patentee may suffer by using functional language in claim limitations that are not subject to section 112(f):



the dissent argued that, under the majority opinion, “one minor drafting decision greatly

expands the scope of the claim limitation” and “patent applicants are able to claim broad

functionality without being subject to the restraints imposed by § 112 ¶ 6.”60

Only a year after Apple, the Federal Circuit acted en banc in Williamson v. Citrix Online

to undermine the primacy of intent theory and elevate scope theory in the threshold test.61

Drawing from the policy concerns about the overbreadth of functional claims that led the

Supreme Court to invalidate functional claims in the first place, the Federal Circuit noted that a

strong presumption against using section 112(f) when a claim limitation did not use the word

“means”:

has the inappropriate practical effect of placing thumb on what should otherwise be a balanced analytical scale. It has shifted the balance struck by Congress in passing § 112, para. 6 and has resulted in a proliferation of functional claiming untethered to § 112, para. 6 and free of the strictures set forth in the statute.62

The Federal Circuit proceeded to flatly state that the “heightened burden” announced in

Lighting World “is unjustified” and that it “should abandon characterizing as ‘strong’ the

We recognize that the choice to draft a claim in ‘broad [functional] terms’ rather than in means-‐plus-‐function format may render the claim more vulnerable to an invalidity attack. Whether to draft a claim in broad [functional] terms is the claim drafter’s choice, and any resulting risk that emanates from that choice is not a basis for the court to rewrite a claim in means-‐plus-‐function format.

Id. at 1298–99; see also Lighting World, Inc. v. Birchwood Lighting, Inc., 382 F.3d 1354, 1361–62 (Fed. Cir. 2004). This passage seems to recognize that claims drafted with “broad functional terms” may be vulnerable to attack under contemporary validity doctrines like enablement and written description, but it seems to ignore the Supreme Court’s cases indicating that claims drafted with “broad functional terms” are facially invalid. See supra __. 60 Id. at 1337 (Prost, J., concurring-‐in-‐part and dissenting-‐in-‐part). 61 Williamson v. Citrix Online, LLC, 792 F.3d 1339 (Fed. Cir. 2015) (en banc in part). Only the section addressing the proper standard for the 112(f) threshold was authored en banc. 62 Id. at 1349.



presumption that a limitation lacking the word ‘means’ is not subject to § 112 ¶ 6.”63 Instead,

“[t]he standard is whether the words of the claim are understood by persons of ordinary skill in

the art to have sufficiently definite meaning as the name for structure.”64

However, even Williamson does not represent a complete triumph of scope theory over

intent theory. The Federal Circuit did not abandon the use of presumptions in the section 112(f)

threshold test. The presumption that section 112(f) does not apply to limitations without the

term “means” persists, and Williamson only made it easier to rebut,65 and the converse, pre-‐

Williams strong presumption that the use of the word “means” does trigger section 112(f)

remains in place.66

B. Uncertainties in the Quantitative Question after Williamson

One important question after the Federal Circuit’s tilt toward scope theory in

Williamson will be the quantitative question of the threshold test: How much structure is

enough to avoid section 112(f)?67 Four unresolved issues that all contribute to uncertainty in

the answer to this question after Williamson are noted below.

First, as a Williamson concurrence notes, the Federal Circuit does not clarify what force

the not-‐strong presumption against section 112(f) that follows from the absence of “means” 63 Id. Judge Newman dissented on this point. Channeling intent theory, she argued that “it is the applicant’s choice during prosecution whether or not to invoke paragraph 6, and the court’s job is to hold the patentee to his or her choice.” Id. at 1358 (Newman, J., dissenting). 64 Id. at 1349. 65 Id. In the concurring section of his concurring-‐in-‐part and dissenting-‐in-‐part opinion, Judge Reyna argues that even this weaker presumption should be abandoned in favor of a more flexible framework. Id. at 1356–57. 66 Id. at 1349. 67 Michael Risch, The Past and Future of Functional Claiming …, Written Description (June 16, 2015), available at http://writtendescription.blogspot.com/2015/06/the-‐past-‐and-‐future-‐of-‐functional.html.



has after Williamson.68 Earlier Federal Circuit opinions suggest that the burden of persuasion is

a preponderance of the evidence,69 but whether Williamson returns to this burden is unclear.

Second, future cases will have to flush out the substantive standard. Under what

conditions does the claim limitation have “sufficiently definite meaning as the name for

structure”? Any old bit of structure will likely not be enough. In fact, Williamson picks up on

earlier case law suggesting that this sufficient-‐structure standard may not be unitary but may

instead have two different prongs: “When a claim term lacks the word ‘means,’ the

presumption can be overcome and § 112, para. 6 will apply if the challenger demonstrates that

the claim term fails to ‘recite sufficiently definite structure’ or else recites ‘function without

reciting sufficient structure for performing that function.’”70 Are these two requirements

merely distinct verbal formulations of the same standard, or are they substantively different

threshold that must both be satisfied? One post-‐Williamson district court opinion interpreted

them to establish distinct thresholds for sufficient structure and held that the claims at issue

satisfied the first but not the second.71

Third, and potentially most problematically, the Federal Circuit has never clarified the

methodology that should be used to determine whether a claim recites sufficient structure.

Some cases seem to state that the default claim-‐construction methodology specified in Phillips

68 Williamson, 792 F.3d at 1357 n.2 (Reyna, J., concurring-‐in-‐part and dissenting-‐in-‐part). 69 Apex Inc. v. Raritan Computer, Inc. 325 F.3d 1364, 1372 (Fed. Cir. 2003). 70 Williamson, 792 F.3d at 1349 (quoting Watts v. XL Sys., Inc., 232 F.3d 877, 880 (Fed.Cir.2000)) (emphasis added); see also CCS Fitness, 288 F.3d 1359, 1369 (Fed. Cir. 2002). 71 The district court concluded that the claim limitation “processor for associating the content data with dispatch record data” recited both “sufficiently definite structure” and “function without sufficient structure for performing that function.”). GoDaddy.com, LLC v. RPost Comm’ns Ltd., 2016 WL 212676, at *53–*57 (D. Ariz. Jan. 19, 2016).



v. AWH72 should control the search for sufficient structure.73 The use of Phillips would make

sense as a policy matter. The Supreme Court’s concerns about overbroad functional claims are

only mitigated when the scope of the claim in infringement proceedings has sufficient

structural limitations, and Phillips controls claim scope for infringement purposes. However, the

Federal Circuit often speaks of the search for sufficient structure in the threshold test not in the

exact terms outlined in Phillips but rather in different terms that roughly mirror Phillips. For

example, Federal Circuit cases don’t usually ask whether a claim limitation has a “meaning” that

is sufficiently structural, which is the usual way of talking about claim construction, but rather

whether a limitation “connotes” sufficient structure.74 Is connotation the same as meaning, or

can a limitation connote structure (in the sense of vaguely invoking it) without that structure

actually being a limit on literal claim scope under Phillips? Similarly, Federal Circuit cases are

not precise about the hierarchy in authority of the possible interpretive sources that can be

used to determine whether a claim limitation connotes sufficient structure. It is well

established that the Federal Circuit performs the section 112(f) threshold test “on an element-‐

by-‐element basis, based upon the patent and its prosecution history.”75 In some cases, as under

Phillips, the structure comes from extrinsic evidence such as trade usage by the person having

72 Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc). 73 Personalized Media Communications, LLC v. U.S. Int’l Trade Comm’n, 161 F.3d 696, 704 (Fed. Cir. 1998) (stating that, in the section 112(f) threshold test, “the focus remains on whether the claim as properly construed recites sufficiently definite structure to avoid the ambit of § 112, ¶ 6”). 74 See, e.g., Apex v. Raritan Computer, Inc., 325 F.3d 1364, 1373 (Fed. Cir. 2003) (“The threshold issue for all the limitations involving the term “circuit” is whether the term itself connotes sufficient structure to one of ordinary skill in the art to perform the functions identified by each limitation.”) 75 Cole v. Kimberly-‐Clark Corp., 102 F.3d 524, 531 (Fed. Cir. 1996).



ordinary skill.76 In other cases, again as under Phillips, the structure comes from intrinsic

sources. Language accompanying the functional term in the limitation itself may provide the

structure,77 or the claim language may have structure when it is read in the context of the

specification.78 However, the key to Phillips and the Federal Circuit’s claim construction

doctrine as a whole is the guidance that it gives about which of these interpretive sources

should be given greater weight.79 To date, this discussion about the balance of the different

interpretive sources has been absent in the search for sufficient structure during the threshold

test. In some cases, the Federal Circuit seems far more willing to read structure from the

specification into the claims as part of the search for sufficient structure in the threshold test

than it would be to read limiting features of preferred embodiments into the claim limitations

under Phillips.80

Fourth, it is unclear what constitutes a question of law and a question of fact in the

threshold test. In Teva Pharmaceuticals v. Sandoz Inc., the Supreme Court held that appellate

76 Greenberg v. Ethicon Endo-‐Surgery, Inc., 91 F.3d 1580, 1583 (Fed. Cir. 1996) (holding that although its plain dictionary meaning was expressed in functional language, “as the name for structure, has a reasonably well understood meaning in the art”) 77 Watts v. XL Sys., Inc., 232 F.3d 877, 880–81 (Fed. Cir. 2000) (holding that a “sealingly connected” limitation was not governed by section 112(f) because it also recited “a second end formed with tapered external threads”); Al-‐Site Corp. v. VSI Int’l, Inc. 174 F.3d 1308, 1318 (Fed. Cir. 1999) (holding that a claim reciting a “member” for “mounting a pair of eyeglasses” was not a section 112(f) claim because, “the claims themselves contain sufficient structural limitations for performing” that function, such as the requirement that the eyeglass hanger member be “made from flat sheet material” with an “opening means formed ... below [its] upper edge”). 78 Inventio AG v. Thyssen Krupp Elevator Americas Corp., 649 F.3d 1350, 1357–59 (Fed. Cir. 2011) (giving significant weight to the written description in determining that a “modernizing device” limitation recited sufficient structure so that it was not governed by section 112(f)). 79 There have been complaints about the clarity of the guidance given in Phillips. See, e.g., Retractable Techs., Inc. v. Becton, Dickinson & Co., 659 F.3d 1369 (Fed. Cir. 2011) (Moore, J., dissenting from the denial of the petition for rehearing en banc) (“Retractable simply cannot be reconciled with our en banc decision in Phillips.”). 80 See, e.g., infra __ (discussing Apple v. Motorola). This willingness to rely on the specification to identify sufficient structure may have resulted in part from the strong presumption against the invocation of section 112(f) when the claim limitation does not employ the term “means,” so it is unclear whether it will continue after Williamson.



courts cannot review district courts’ claim construction rulings on an entirely de novo basis but

must instead review any underlying issues of fact decided by the district court more

deferentially under a clear error standard of review.81 The section 112(f) threshold test is an

issue of claim construction and thus ultimately a question of law,82 but there are likely to be

factual findings, such as the trade usage of a term to a person of ordinary skill, that are wound

up in the decision. In Lighting Ballast Control v. Phillips Electronics, the Federal Circuit deferred

to a district court’s determination that the claim term “voltage source means” was not a section

112(f) limitation based on trade usage: “it is … understood by persons of skill in the art to

connote a class of structures.”83 While trade usage is a question of fact, the quantitative

question—that is, whether there is sufficient structure in that usage to overcome the

presumption in favor of section 112(f) when a claim employs the term “means”—would seem

to be a question of law on which no deference is due.

III. Section 112(f) and Software

The application of section 112(f) to software requires a sui generis modification of

conventional section 112(f) doctrine. Section III.A posits that this modification is required

because software is, for practical purposes at least, a purely functional technology without any

physical structure that is relevant to what a software inventor invents. Section III.B explores

how the Federal Circuit had already adapted section 112(f) to software before Williamson by

81 Teva Pharms. v. Sandoz Inc., 135 S.Ct. 831 (2015). 82 Markman v. Westview Instruments, Inc., 517 U.S. 370 (1996). 83 Lighting Ballast Control LLC v. Philips Electronics North America Corp., 790 F.3d 1329, 1338–39 (Fed. Cir. 2015).



concluding that an “algorithm” is the corresponding structure in the specification for functional

software limitations.

A. The Software/Section 112(f) Mismatch84

There is a fundamental mismatch between the rules of 112(f) and software inventions:

in most technologies, 112(f) builds on an intuitive distinction between the physical structure

and the function of a technology, but, in software, there is no such distinction. Software

inventions are, at least as a practical matter and for the purpose of patent law, a purely

functional technology. To bring section 112(f) to bear on software claims, courts must identify

software’s logical structure, i.e., the type of functional description of a software invention that

should count as metaphorical structure.

In the historical core of section 112(f) claiming in the mechanical arts, what constitutes a

structural property of an invention and how an invention’s structural properties differ from its

functional properties are self-‐evident, intuitive concepts. In brief, a structural property is what

an invention “is,” whereas a functional property is what an invention “does.”85 For example, a

“skid plate,”86 “button and hole arrangement,”87 and “a rotatable disc with an opening sits

above the container cap”88 are all structural descriptions of objects.89 Few—if any—cases in the

84 Much of the argument in this section draws from Collins, supra note __, at 1440–43. 85 In re Swinehart, 439 F.2d 210, 212 (C.C.P.A. 1971). 86 Chiuminatta Concrete Concepts, Inc. v. Cardinal Industries, Inc., 145 F.3d 1303, 1309 (Fed. Cir. 1998) (noting that the specification described a skid plate as “a generally rectangular strip of metal having rounded ends … between which is a flat piece” and further specified that “[t]he flat piece … is generally parallel to the base plate”). 87 Al-‐Site Corp. v. VSI International, Inc., 174 F.3d 1308, 1315–16 (Fed. Cir. 1999). 88 Sage Products, Inc. v. Devon Industries, Inc., 126 F.3d 1420, 1428 (Fed. Cir. 1997). Here, “rotatable” adds a dollop of function to the description.



mechanical arts bother even to expressly pose the definitional “What is structure?” question

because the answer is so self-‐evident. To be clear, there are plenty of cases in the mechanical

arts in which it is unclear whether a term recites any structure as a limitation on claim scope or

whether the quantum of structure recited is sufficient.90 However, most of this uncertainty

inheres in either the meaning of language or the legal doctrine that controls the answer to

quantitative question, not the nature of a structural property of a thing. That is, the uncertainty

follows from whether claim language refers to enough of the structural properties of a

technology, not whether any particular property, once identified as a claim limitation, is a

structural property.91

In the software arts, however, what should count as structure for the purposes of

section 112(f) is not intuitive. Three links in a chain of arguments are important to emphasize

here. First, software programs do have physical, structural properties, just like other

technologies do. However, second, those physical, structural properties are of no use in section

112(f). They are irrelevant to the definition of a software invention because software inventions

are functional entities on all relevant levels of definition. Thus, third, the definitional question—

“What is structure in the software arts?”—must focus on logical structure. It must identify a

functional definition of a software invention that is sufficiently granular to serve as 89 Structure is also an uncontroversial concept in the chemical and biochemical arts because structure is equated with molecular structure. Ariad Pharm., Inc. v. Eli Lilly & Co., 598 F.3d 1336 (Fed. Cir. 2010) (en banc). However, the task of policing overbroad, functional claims in these arts has fallen to the written description doctrine rather than to means-‐plus-‐function claiming. Collins, supra, note __, at 1430–33. 90 See supra __ (discussing the quantitative question of the section 112(f) threshold test). 91 The distinction between structural and functional properties is a sufficiently basic and intuitive that it often underlies the philosophical analysis of the properties of things themselves. See, e.g., Peter Kroes, Technological Explanations: The Relation Between Structure and Function of Technological Objects, 3 Phil. & Tech. 18, 18 (1998) (discussing “two different modes of description, viz., a structural and a functional mode of description” for technological objects).



metaphorical structure, at least if the scope-‐narrowing strictures of section 112(f) are to be

brought to bear on software at all.

First, despite the commonplace description of software as “intangible,”92 every real-‐

world embodiment of a software program is a material, worldly entity that has structural

properties, just like every embodiment of a mechanical technology has structural properties.93

Software exists as electrons or charges on a hard drive or in a computer’s memory; a computer

implements a software program only because a particular set of gates or switches in the

processor is open or closed.94 “The instructions of the software program … electrically change

the general purpose computer by creating electrical paths within the device.”95

Second, the physical, structural, and material properties of a software embodiment are

usually irrelevant to the task of identifying, delineating, or defining a protectable software

invention, whether under section 112(f) or any other patent doctrine that invokes structure. In

92 See, e.g., Bancorp Servs. v. Sun Life Assurance Co., 687 F.3d 1266, 1277–79 (Fed. Cir. 2012) (making an analogy between computer-‐executed and and mental processes); In re Grams, 888 F.2d 835 (Fed. Cir. 1989) (same); In re Bernhart, 417 F.2d 1395 (C.C.P.A. 1969) (same); Richard S. Gruner, Intangible Inventions: Patentable Subject Matter for an Information Age, 35 LOY. L.A. L. REV. 355, 357 (2002) (“New designs for software and computer-‐based business practices . . . . resemble the sorts of intangible ideas and thought processes that have traditionally fallen outside of patent protections.”). 93 Microsoft Corp. v. AT&T Corp., 550 U.S. 437 (2007) (discussing the physicality of any copy of a software program that can generate functional effects in the course of assessing when software can be a “component” under § 271(f)); Robert Plotkin, Computer Programming and the Automation of Invention: A Case for Software Patent Reform, 7 UCLA J.L. & TECH. 1, 8–12 (2003). 94 WMS gaming, at 1348 & n.3; Plotkin, supra note __, at 38–39. In fact, patent law has unflinchingly recognized the tangibility of software embodiments in some of its doctrines even as it has denied the tangibility of software in others. For example, a computer programmed with a new software program has long been treated as a new machine under the novelty doctrine that is structurally distinct from prior art machines. In re Alappat, 33 F.3d 1526, 1545 (Fed. Cir. 1994) (en banc); In re Freeman, 573 F.2d 1237, 1247 (C.C.P.A. 1978); In re Bernhart, 417 Fe.2d 1395, 1399–1400 (C.C.P.A. 1969); In re Prater, 415 F.2d 1393, 1403 n. 29 (C.C.P.A. 1969). 95 WMS Gaming Inc. v. Int’l Game Tech., 184 F.3d 1339, 1349 (Fed. Cir. 1999).



most arts, it is natural to say that an invention is its structure, not its function.96 In the software

arts, however, the opposite characterization seems more natural: an invention is its function,

not its structure.97 A usable embodiment of a software invention may not be literally

immaterial, but the description of a software invention must be nothing but function as a

practical matter. The irrelevance of the physical, structural properties of a software

embodiment to the definition of a software program has been engineered into the very nature

of software itself. The core value of software lies in the fact that the design of software that

possesses any given set of logical, functional properties need not involve any consideration of

either the physical properties of the hardware or the distribution of electrical charges on a disk.

In fact, “[F]or all practical purposes the programmer and others who think about and describe

the program have no practical choice but to conceive of and describe it in terms of its logical

structure [or function]. . . . It is far from clear that it would even be possible for the human mind

to appreciate the physical structure of all but the simplest programs or to explain them in terms

of their physical structures.”98 A software program can be implemented in entirely different

code in the same programming language or in entirely different languages, and the sets of code

96 See supra __. 97 Note, Everlasting Software, 125 Harv. L. Rev. 1454, 1456 (2012). 98 Plotkin, supra note __, at 46 & n.126; see also id. at 26 (“The process of computer programming enables a programmer to create a machine that has a particular novel physical structure for performing a particular function without requiring the programmer to design the novel features of the machine in physical terms.”); id. at 36 (“[O]ne of computer science’s express goals is to ensure that programmers can do their work in complete ignorance of the physical structure of . . . hardware . . . .”); id. at 44–45 (“[A] programmer who modifies the physical structure of a computer by providing source code to the computer need not even know that the computer’s memory is being physically modified at all, much less understand or appreciate the nature of those physical modifications.”) (citations omitted).



that all embody a software invention have few, if any, structural properties in common.99

Thanks to interpreters and compilers, any given program can be implemented on a wide array

of different computers, each possessing a different internal architecture and requiring the

software to take on different physical, structural properties.100 In sum, the many distinct

embodiments of a software invention are unlikely to have any physical, structural properties in

common that can be used to delineate what an inventor has invented. For patent drafters

writing software claims, software functionality is like a never-‐ending set of nested Russian dolls:

you open up one more general functional description to look for structure, and all you find is

another, more specific functional description.101

Third, the purely functional nature of software raises a challenge for any patent doctrine

that builds on the categorical distinction between the structural and functional properties of an

invention and prohibits purely functional claim limitations. This mismatch between software

and section 112(f) presents two doctrinally simple solutions, but both are so extreme that

neither is palatable.102 On the one hand, we could refuse to make any allowances for the purely

functional nature of software and mandate that software patents recite physical, structural

properties as limitations on claim scope. This option would render patent protection for

software absurd and economically meaningless because avoiding infringement would be a

99 Furthermore, hardware and software implementations of any given program are functionally interchangeable despite their radically different structural properties. In re Alappat, 33 F.3d 1526, 1583 (Fed. Cir. 1994) (Rader, J., concurring). 100 See W. DANIEL HILLIS, THE PATTERN ON THE STONE, at 56–58 (1998) (discussing interpreters and compilers); cf. Microsoft Corp. v. AT&T Corp., 550 U.S. 437, 450 (2007) (“Software . . . is a stand-‐alone product developed and marketed ‘for use on many different types of computer hardware . . . .’”). 101 Technically, the nested dolls do end at some point. See infra __ (discussing In re Katz). 102 If either were adopted, it might well be preferable to eliminate patent protection for software altogether.



trivial undertaking. On the other hand, we could exempt software from the strictures of section

112(f) and permit purely functional claiming at any level of generality. This option ignores the

Supreme Court’s cases from the early twentieth century that prohibit purely functional claiming

and raises significant policy concerns about the social costs of software-‐claim overbreadth.103

Given the problematic nature of both simple, but extreme, doctrinal solutions, an exploration

of the sui generis ways in which section 112(f) could be modified in the purely functional arts

like software is a worthwhile undertaking to explore.104 Such an exploration of when function

should count as metaphorical structure offers the greatest promise for way to use section

112(f) to navigate a course between the Scylla and Charybdis of trivially narrow and vastly

overbroad claims.

B. Algorithms as Corresponding Structure

The need to create a sui generis doctrine for section 112(f) to deal with software claims

has not been lost on the Federal Circuit. The Federal Circuit has developed an extensive body of

cases addressing what constitutes the corresponding structure of a section 112(f) software

limitation in a patent disclosure.105

Initially, it is important to emphasize that the Federal Circuit has rejected the idea that

corresponding structure for the purposes of section 112(f) is physical structure. The Federal

Circuit has repeatedly held that the disclosure of a general purpose computer or generic

103 See supra __. 104 The jury is still out, however, on whether it will succeed. See infra __. 105 Some pre-‐Williamosn Federal Circuit cases also address what constitutes structure in a claim limitation as part of the section 112(f) threshold test, but the reasoning in these cases is either misguided (because it relies on physical structure) or radically underdeveloped. See infra Section IV.B.



microprocessor in the specification that can execute software is not the corresponding

structure for software limitations.106 Thus, as the Federal Circuit recently stated, “‘[s]tructure’

to a person of ordinary skill in the art of computer-‐implemented inventions may differ from

more traditional, mechanical structure …. Indeed, the typical physical structure that

implements software, a computer, cannot be relied upon to provide sufficiently definite

structure for a software claim ….”107 Nor has the Federal Circuit even entertained the notion

that the disclosure of a distribution of electronics on a storage medium or an arrangement of

open and closed gates in a computer, which are the structural properties of a software

invention that differentiate the invention from all other software programs, should be the

corresponding structure for a software limitation.

Rather, the Federal Circuit has held that the corresponding structure in the specification

for the purposes of section 112(f) is metaphorical structure—a technically functional

description of software that can play the scope-‐limiting role that physical structure plays in the

mechanical arts. This adoption of metaphorical structure in the software arts means that we

are talking about different ideas altogether when we discuss “structure” for the purpose of

section 112(f) and “structure” for the purpose of the section 101 prohibition of In re Nuijten on

106 See, e.g., EON Corp. v. AT&T Mobility LLC, 785 F.3d 616, 621 (Fed. Cir. 2015); Aristocrat Techs. Austl. Pty Ltd. V. Int’l Game Tech., 521 F.3d 1328, 1333 (Fed. Cir. 2008); WMS Gaming Inc. v. Int’l Game Tech., 184 F.3d 1339, 1349 (Fed. Cir. 1999). But see infra __ (discussing the Katz exception). Technically, a microprocessor is a structural feature of the claimed technology. However, it should not be enough structure to serve as corresponding structure. The analogy in the mechanical arts to the disclosure of a microprocessor as corresponding structure might be the disclosure of a material like metal as the corresponding structure. Being made of metal is a structural property, but it is not enough information about structure to say that the specification has disclosed corresponding structure. 107 Apple Inc. v. Motorola Inc., 757 F.3d 1286, 1298 (2014). In this same passage, the opinion also states that “looking for traditional ‘physical structure’ in a computer software claim is fruitless because software does not contain physical structures.” Id. This statement is technically inaccurate. See supra __.



claims to propagating signals or software per se.108 Thus, there is ambiguity in the term

“structure” in the law of software patents. The latter definition that is relevant to section 101

requires physical structure, and recitation of a microprocessor or a fixed storage medium is

enough to avoid a patent-‐ineligible claim to software per se. However, the former definition

that is relevant to section 112(f) does not refer to physical structure.

The Federal Circuit’s answer to the definitional “What is structure?” question in the

context of searching for corresponding structure in the specification is that structure is an

“algorithm.” What the Federal Circuit means by an algorithm is “in essence a series of

instructions for the computer to follow,”109 “a step-‐by-‐step procedure for accomplishing a given

result,”110 or a “sequence of computational steps to follow.”111 Simply put, “the essence of

algorithms” is “what to do to perform a task” that is recited as a claim limitation.112

The Federal Circuit first introduced the notion of an algorithm as the corresponding

structure in its 1999 opinion in WMS Gaming, Inc. v. International Game Technologies.113 The

claim limitations at issue used the word “means,” and the parties did not dispute that the claim

terms at issue were subject to section 112(f) under the threshold test. The district court

construed the limitation broadly to encompass any table, formula, or algorithm for performing 108 In re Nuijten, 500 F.3d 1346, 1356 (Fed. Cir. 2007) (“While such a signal is physical and real, it does not possess concrete structure ….”); id. n.6 (noting that the signal on a “storage medium” would be patent eligible). 109 Typhoon Touch Techs., Inc. v. Dell, Inc., 659 F.3d 1376, 1384 (Fed. Cir. 2011) (quoting In re Waldbaum, 457 F.2d 997, 998 (C.C.P.A. 1972)). 110 Id. The PTO defines an algorithm as a “finite sequence of steps for solving a logical or mathematical problem or performing a task.” [PTO Guidelines, quoting Microsoft Computer Dictionary, Microsoft Press, 5th edition 2002.] 111 Ibormeith IP, LLC 14 v. Mercedes-‐Benz USA, LLC, 732 F.3d 1376, 1379 (Fed. Cir. 2013) (citations omitted). 112 Allen Newell, Response: The Models Are Broken, The Models Are Broken, 47 U. PITT. L. REV. 1023, 1024 (1986); see also id. at 1026 (“An algorithm is [a] . . . sequence of steps or operations for solving a class of problems.”). 113 WMS Gaming Inc. v. Int’l Game Tech., 184 F.3d 1339, 1347–50 (Fed. Cir. 1999).



the claimed function, but the Federal Circuit insisted that the scope of the claim was narrower

because it was limited to “the special purpose computer programmed to perform the disclosed

algorithm” in the specification, as well as equivalents.114 In subsequent cases, the Federal

Circuit has clarified that there is no single format in which an algorithm must be communicated.

Mathematical formulae, prose, and flow charts can all express algorithms.115

Starting in the late 2000s, the Federal Circuit began to invalidate a large number of

software claims with section 112(f) limitations for indefiniteness because the specification did

not disclose an algorithm as corresponding structure.116 These cases offer an interesting

window into what constitutes an algorithm. If the specification simply repeats the functions

recited in the claims, the specification does not disclose an algorithm. For example, in Finisar

Corp. v. Direct TV Group, Inc., a claim recited the section 112(f) limitation “database editing

means … for generating a hierarchically arranged set of indices for referencing data in said

information database.”117 The Federal Circuit affirmed the district court’s holding of

indefiniteness because the specification provided “nothing more than a restatement of the

114 Id. at 1349. 115 Finisar Corp. v. DirecTV Grp., Inc., 523 F.3d 1323, 1340 (Fed. Cir. 2008). 116 Eon Corp. IP Holdings LLC v. AT&T Mobility LLC, 785 F.3d 616 (Fed. Cir. 2015); Triton Tech, LLC v. Nintendo, Inc., 753 F.3d 1375 (Fed. Cir. 2014); Robert Bosch, LLC v. Snap-‐On Inc., 769 F.3d 1094 (Fed. Cir. 2014); Function Media, L.L.C. v. Google Inc., No. 2012-‐1020, 2013 WL 516366 (Fed. Cir. Feb. 13, 2013); ePlus, Inc. v. Lawson Software, Inc., 700 F.3d 509 (Fed. Cir. 2012); Noah Sys., Inc., v. Intuit Inc., 675 F.3d 1302 (Fed. Cir. 2012); HTC Corp. v. IPCom GmbH & Co., 667 F.3d 1270 (Fed. Cir. 2012); Dealertrack, Inc. v. Huber, 674 F.3d 1315 (Fed. Cir. 2012); In re Aoyama, 656 F.3d 1293 (Fed. Cir. 2011); Rembrandt Data Techs. v. AOL, 641 F.3d 1331 (Fed. Cir. 2011); In re Katz Interactive Call Processing Patent Litig., 639 F.3d 1303 (Fed. Cir. 2011); Brown v. Baylor Healthcare Sys., No. 2009-‐1530, 2010 WL 1838921 (Fed. Cir. May 7, 2010); Net MoneyIN, Inc. v. VeriSign, Inc., 545 F.3d 1359, 1366–67 (Fed. Cir. 2009); Encyclopaedia Britannica, Inc. v. Alpine Elecs., Inc., No. 2009-‐1087, 2009 WL 4458527 (Fed. Cir. Dec. 4, 2009); Blackboard, Inc. v. Desire2Learn Inc., 574 F.3d 1371, 1383–85 (Fed. Cir. 2008); Finisar Corp. v. DirecTV Grp., Inc., 523 F.3d 1323 (Fed. Cir. 2008); Aristocrat Techs. Austl. Pty. Ltd. v. Int’l Game Tech., 521 F.3d 1328 (Fed. Cir. 2008). 117 Finisar Corp. v. Direct TV Group, Inc., 523 F.3d 1323, 1340 (Fed. Cir. 2008).



function, as recited in the claim.”118 However, if the specification provides a more granular

functional description—that is, if it discloses a series of functional steps that, when strung

together, achieve the claimed function—then the specification does disclose an algorithm. For

example, in Typhoon Touch Techs. v. Dell, a claim recited the section 112(f) limitation “means

for cross-‐referencing said responses with one of said libraries of said possible responses.”119

The district court held that there was no corresponding algorithm disclosed in the specification

and that the claims containing the limitation were invalid for indefiniteness. The Federal Circuit

reversed, holding that the specification “recite[d] a four-‐step algorithm for computer-‐

implemented cross-‐referencing, starting with the entry of a response, then a search for the

entered response in a library of responses, then determination whether a match exists in the

library, and then execution of an action if a match exists.”120 The specification described a series

of four functional tasks, each of which was more granular than the functional task recited as a

claim limitation. When strung together, these more granular tasks formed a process for

accomplishing the claimed task. They were thus an algorithm, and they qualified as

corresponding structure.

The Federal Circuit has identified one category of functional limitations in software

claims that do not require the disclosure of any corresponding structure. In In re Katz

Interactive Call Processing Patent Litigation, the Federal Circuit held that the claimed functions

of “processing,” “receiving,” and “storing” did not require the disclosure of an algorithm even

118 Id. 119 Typhoon Touch Techs., Inc. v. Dell, Inc., 659 F.3d 1376, 1383 (Fed. Cir. 2011). 120 Id. at 1385.



though the limitations were subject to section 112(f).121 It reasoned that the disclosure of an

algorithm was needed to have corresponding structure when the claims described “specific

functions that would need to be implemented by programming a general purpose computer to

convert it into a special purpose computer capable of performing those specified functions.”122

However, it concluded that the claimed functions at issue in Katz could rely on the disclosure of

a general purpose computer as the corresponding structure because they “can be achieved by

any general purpose computer without special programming. As such it was not necessary to

disclose more structure than the general purpose processor that performs those functions.”123

To reiterate, apart from the “narrow” Katz exception,124 section 112(f) software

limitations require the disclosure of corresponding structure in the form of an algorithm or a

step-‐by-‐step procedure for achieving the claimed functionality. However, an algorithm does not

disclose physical structure. The steps that make up an algorithm are still functional descriptions

of tasks to be performed. For example, each of the steps in the Typhoon Touch Techs.

algorithm—such as the step of “search[ing] for the entered response in a library of

responses”—is just as functional as the claimed task—namely “cross-‐referencing said responses

with one of said libraries of said possible responses.”125 The difference is only the level of

granularity of the functional description. The Federal Circuit does occasionally imply that an

121 In re Katz Interactive Call Processing Patent Litigation, 639 F.3d 1303, 1316 (Fed. Cir. 2011). 122 Id. 123 Id. 124 Subsequent Federal Circuit cases that have considered the Katz exception have called it “narrow,” have not found it to apply, and have required the disclosure of an algorithm in the specification to avoid indefiniteness. EON Corp. v. AT&T Mobility LLC, 785 F.3d 616, 621–23 (Fed. Cir. 2015); Ergo Licensing, LLC v. CareFusion 303, Inc., 673 F.3d 1361, 1364 (Fed. Cir. 2012). 125 See supra __.



algorithm is something different from a functional description of software by juxtaposing an

algorithm to functional language. For example, it explained its invalidation of a software claim

with a section 112(f) limitation for indefiniteness with the following reasoning: “The

specification merely provides functional language and does not contain any step-‐by-‐step

process for controlling the adjusting means.”126 This statement is misleading because it implies

that an algorithm is not expressed with functional language. What the Federal Circuit means by

this statement and other similar ones is, more precisely, that the purported algorithm cannot

merely provide functional language at the same level of generality as the functional language

that forms the claim limitation. Occasional loose judicial descriptions notwithstanding, an

algorithm is a functional entity that relates metaphorical, not physical, structure.

IV. The Definitional Question after Williamson: Pre-‐Williamson Doctrine Is Unhelpful

At first glance, Williamson might not appear to call for a revolution in the section 112(f)

threshold test in the software arts. While Williamson’s turn to scope theory will force the

Federal Circuit to ask and answer the definitional “What is structure?” question as part of the

threshold test on a more frequent and open basis, the Federal Circuit was, in theory at least,

already identifying software’s structure in section 112(f) cases in two distinct doctrinal contexts

even before Williamson. First, it was entertaining arguments about the presence or absence of

structure in claim limitations that were intended to rebut the pre-‐Williamson, strong

presumptions of the threshold test. Second, after limitations had been labeled as section 112(f)

126 Ergo Licensing, LLC v. CareFusion 303, Inc., 673 F.3d 1361, 1365 (Fed. Cir. 2012).



limitations, it was looking for corresponding structure in the specification in the form of

algorithms.

However, Williamson will in fact require a radical shift in the Federal Circuit’s section

112(f) software doctrine. Upon closer examination, pre-‐Williamson cases turn out to provide

surprisingly little guidance for answering the definitional question as part of the threshold test.

Starting with the better-‐developed body of law, Section IV.A demonstrates that the Federal

Circuit’s existing doctrine identifying algorithms as software’s structure cannot simply be

portaged from the search for corresponding structure in a specification to the search for

structure in a claim limitation. The principal problem is that the Federal Circuit’s definition of an

algorithm only defines structure in relation to the baseline of a particular claim, and there is no

baseline in relation to which one can assess whether a claim itself recites an algorithm. If the

Federal Circuit were to rely on this relative definition of structure to identify structure in the

threshold test, all software limitations would become section 112(f) limitations.127 Section IV.B

then looks for crumbs of wisdom in the Federal Circuit’s pre-‐Williamson cases that address the

definitional question as part of the threshold test. Because the strong presumptions did most of

the work, opinions addressing this issue are generally of little help in the post-‐Williamson

world, but one line of reasoning does hint at a protean test for identifying a stand-‐alone

definition of structure—that is, a definition that does not rely on a comparison to the claims as

a baseline.

127 Alternatively, the Federal Circuit could conclude that no software limitations are subject to section 112(f), but this outcome is unlikely. See infra __.



A. Algorithms as Corresponding Structure as a Model?

The Federal Circuit’s identification of algorithms as corresponding structure in the

specification for section 112(f) software limitations is by now well established.128 One

understandable, initial reaction to Williamson’s impact on software in the patent community

might therefore be a collective shrug of the shoulders.129 If there is a well-‐established answer to

the definitional question in the context of identifying corresponding structure in the

specification, then it might seem reasonable to assume that this answer could be simply carried

over and used to identify structure in the claim limitations.130 Upon closer inspection, however,

this assumption proves to be unfounded. The Federal Circuit’s cases that position algorithms as

corresponding structure for limitations that are already governed by section 112(f) are

unhelpful when the task at hand is determining whether there is sufficient structure in the

limitations to avoid the application of section 112(f) in the first place.

The initial problem is that there is no single, functionally defined claim limitation that is

potentially subject to section 112(f) that can ever be an algorithm, and thus structure, in and of

itself. Any attempt to identify a single claim limitation as an algorithm is premised on a category

error. Functional claim limitations are singular, and algorithms are plural. As the Federal Circuit

has defined the concept, an algorithm is a “sequence of steps for … performing a task.”131 Only

128 See supra Section III.B. 129 [KEC: any commentary that minimizes impact of Williamson?] 130 Cf. Apple Inc. v. Motorola Inc., 757 F.3d 1286, 1294–1304 (2014) (“The overall means-‐plus-‐function analysis is a two-‐step process. Naturally, there is some analytical overlap between these two steps.”). 131 See supra __.



a series of steps, performed one after the other, can be an algorithm. A single-‐step procedure

for performing a task is simply a restatement of the task, not an algorithm.

To carry the Federal Circuit’s algorithm-‐as-‐structure analysis over from the identification

of corresponding structure in a specification to the identification of sufficient structure in a

claim limitation requires a reconfiguration of the usual threshold test. The Federal Circuit

cannot ask whether an individual claim limitation recites structure in the form of an algorithm,

but rather whether a series of strung-‐together functional limitations in a claim constitutes an

algorithm. This approach to identifying structure in the threshold test would require a

significant doctrinal shift, as the threshold test usually examines whether each functional

limitation recites sufficient structure on its own, not whether a series of functional limitations

collectively recites sufficient structure. Each function recited as a claim limitation is usually

analyzed independently to determine whether it has sufficient structure to evade the strictures

of section 112(f), even when the claim states that the same “means” describes more than one

function.132 However, the search for structure in an aggregate of functional limitations is the

only feasible approach if the Federal Circuit’s well-‐established algorithm analysis is to be used

to answer the definitional “What is structure?” question of the threshold test in the software

arts.

Yet, even if one is willing to alter the search for structure in this software-‐specific

manner, the Federal Circuit’s case law on algorithms is decidedly unhelpful in the threshold

test. Section IV.A.1 notes that the Federal Circuit has only defined an algorithm, and thus

132 Media Rights Technologies v. Capital One Financial Corp., 800 F.3d 1366, 1373 (Fed. Cir. 2015); Noah Sys., Inc. v. Intuit Inc., 675 F.3d 1302, 1318–19 (Fed. Cir. 2012).



corresponding structure for software limitations, in a relative manner. Section IV.B.2 then

demonstrates why a relative definition of structure is not up to the task of identifying sufficient

structure as part of the post-‐Williamson threshold test.

1. The Federal Circuit’s Relative Definition of an Algorithm

The Federal Circuit has only defined an algorithm in a relative manner. A functional

description of a software invention in the specification is an algorithm if the specification

describes how the software operates in a manner that is more granular than the baseline

functional description of the software in the claim limitation.133 One cannot identify an

algorithm under the Federal Circuit’s definition without first having identified the baseline task

in the claim that the algorithm is supposed to perform.

The relative nature of the Federal Circuit’s definition of an algorithm is on full display in

the fact that it is impossible to examine a specification in isolation from the claims and

meaningfully opine on whether or not it describes corresponding structure. The exact same

specification could provide corresponding structure for one claim and yet fail to provide

corresponding structure for another claim. For example, consider a hypothetical patent on a

digital rights management technology.134 Assume that the specification states that the

invention performs three functions: it controls a data output of a client system by diverting a

commonly used data pathway of a media player application to a controlled data pathway,

monitors the controlled data pathway, and disrupts the playing of the media content at the

133 See supra notes __ and accompanying text. 134 This technology in this hypothetical, but not the legal analysis, is based on the patent at issue in Media Rights Technologies v. Capital One Financial Corp., 800 F.3d 1366 (Fed. Cir. 2015).



controlled data pathway when the playing of the media file content is outside of the usage

restriction applicable to the media file.135 Is this a functional description of how the invention

works that should count as an algorithm and thus corresponding structure? The answer

depends on the claim at issue. Assume that Claim 1 recites a single, umbrella limitation such as

“a compliance mechanism for ensuring that only media content within the usage restriction

applicable to the media is played.” The specification likely does describe an algorithm for this

claim because it provides a series of more-‐granular steps that explains how to perform the less-‐

granular task recited as the claim limitation. Now, assume that Claim 2 recites three separate

functional limitations that parallel the three disclosed functions. That is, assume that Claim 2

recites separate “mechanism for controlling,” “mechanism for monitoring,” and “mechanism

for disrupting” limitations. Now, the unchanged specification does not describe an algorithm.

Rather, it merely restates the functional tasks that are recited as claim limitations.136 A relative

definition of corresponding structure means that courts must recognize that the mystery of

disappearing structure is a real phenomenon. In sum, the existence of an algorithm, and thus

corresponding structure, is relative. A judge cannot look at a description of software

functionality, standing alone, and reach a conclusion that the description does or does not

describe structure. A judge can only identify structure in relation to the baseline functional

description provided by the claims.

The Federal Circuit’s use of a relative definition of corresponding structure for software

limitations does not map neatly onto the nature of corresponding structure in other

135 Cf. id. at 1369-‐70. 136 See supra __ (noting that a restatement of the functional claim limitations is not an algorithm).



technologies. In the mechanical arts, the answer to the definitional question is never relative or

contingent. Any given property of a mechanical invention is either a structural or a functional

property; a three-‐legged stool is not structure with respect to one claim but function with

respect to another.137 There may be relativity when determining whether any given structure in

the specification can perform the claimed function. It may well be that a three-‐legged stool is

corresponding structure for a “means for supporting” limitation, but not for a “means for

cutting” limitation. However, there is no relativity in what constitutes a structural description of

a technology.

To be clear, despite its sui generis nature, the Federal Circuit’s relative definition of

structure in software is conceptually workable, if a bit awkward, when it is put to work in the

specific context of identifying corresponding structure in the specification.138 This search for

structure is always conducted after a claim limitation has been held to be subject to section

112(f), so the functional description recited in the claim is always available to serve as a

baseline. When searching for corresponding structure, the courts can always ask the

definitional question in a relational manner: Given this particular claim limitation, does the

specification disclose structure? Furthermore, a relative definition of corresponding structure

enables section 112(f) to do important normative work in some software cases. By mandating

that functional claim limitations be curtailed in scope to the disclosed algorithm and its

137 Cf. supra __. 138 However, the relative definition of structure is conceptually bankrupt when it is employed in legal doctrine for identifying whether there is sufficient structure in a claim limitation to evade the strictures of section 112(f). See infra Section IV.A.2.



equivalents, section 112(f) narrows the permissible scope of software limitations drafted in

sweepingly general, functional terms.139

2. A Relative Definition of Structure in the Threshold Test

When searching for corresponding structure in a specification for a section 112(f)

software limitation, a relative definition of structure is adequate because the claim identifies

the relevant task and provides the needed baseline.140 However, when searching for sufficient

structure in a claim limitation as part of the section 112(f) threshold test, a relative definition of

structure is not adequate. There is no baseline task to which to compare the series of functions

recited as claim limitations to determine if they form an algorithm. Is any given set of functional

claim limitations a string of structureless tasks, or should the set count as logical structure

because it details a step-‐by-‐step, algorithmic procedure for achieving the over-‐arching task

defined by the invention’s utility?141 The answer, of course, is that a string of functional

limitations can be either one of these things, depending upon how one chooses to frame the

task at hand. As explored below, the Federal Circuit can use its relative definition of structure,

developed in the context of the search for corresponding structure, to identify sufficient

structure in the claims in only two ways. It can achieve either one of two extreme outcomes: it

can subject no software limitations to section 112(f), or it can subject all software limitations to

139 Of course, it also narrows the permissible scope of software limitations that are drafted in quite specific, functional terms. See infra __. 140 See supra Section IV.A.1. 141 This question assumes that the claim recites more than one functional limitation. If a claim recites only one functional limitation, that limitation would automatically be a structureless task if the Federal Circuit’s definition of an algorithm were to govern the search for structure in the section 112(f) threshold test. Cf. supra notes __ and accompanying text.



section 112(f). What the relative definition of structure does not permit, however, is for the

Federal Circuit to toggle between these two outcomes in a principled fashion.

On the one hand, perhaps all software limitations are steps in an algorithm for

accomplishing the task of providing the end user of the patented technology with some

utility.142 Following this approach, all functional software limitations would recite sufficient

structure after Williamson, and no software limitations, excepting perhaps those that employ

the term “means” and that are still subject to the strong presumption,143 would be governed by

section 112(f). This outcome would be problematic in light of the Federal Circuits expectation

stated in Williamson that Williamson’s tilt toward scope theory will reduce the formalism of the

threshold test and increase the number of software limitations subject to section 112(f).144

On the other hand, perhaps all functional software limitations refer to structureless

tasks. Following this approach, no functional software limitations connote sufficient structure,

so all such limitations would be subject to the scope-‐narrowing rules of section 112(f),

regardless of whether they use the term “means.” This route forward would satisfy the

expectation that Williamson will increase the number of software limitations subject to section

112(f), and it would do so in spades. In fact, this route forward means that Williamson would

have triggered a revolution in the law of software patents. Williamson would have caused a

tidal shift in the law of software patents from most functional software limitations not being

subject to section 112(f) to all such limitations being subject to section 112(f).

142 Again, claims with only a single functional limitation would never recite structure under this approach. See supra note __. 143 See supra __. 144 See supra __.



Before moving on, it is worth pausing for a moment to recognize that, while using a

relative definition of structure in the section 112(f) threshold analysis may yield an

administrable rule that reduces the permissible scope of software claims from their pre-‐

Williamson norm, it is awkward in another way. It places a formalistic, rather than substantive,

limit on permissible patent scope.145 A relative definition of structure does not cap permissible

claim scope at any particular level of generality. Rather, it mandates that every functional claim

limitation be narrowed to a disclosed algorithm and its equivalents, regardless of the level of

specificity at which those limitations are drafted.146 If a first claim limitation recites function A

and the specification recites algorithmic steps 1, 2, and 3 for function A, the claim containing

the limitation is valid only if limitation A is restricted to the “structure” of steps 1, 2, and 3, as

well as its equivalents. However, if a second claim were to directly recite functions 1, 2, and 3—

which are identical to steps 1, 2, and 3—as separate limitations, section 112(f) would require

that each limitation in this claim, too, be restricted in scope to more granular, algorithmic steps

(or sub-‐algorithmic steps, depending on your perspective) disclosed in the specification. For

example, the limitation reciting function 1 would need to be restricted to the “structure” of

steps 1a, 1b, and 1c.147 The relative nature of the Federal Circuit’s definition of structure is

again on display: every set of functions that are corresponding structure when they are recited

as the steps of an algorithm in the specification cease to be structure when they are recited as a

series of claim limitations. The disappearance of the structural nature of a description of a

145 This is awkward in part because of the goals of Williamson was arguably to reduce the formalism of section 112(f) by shifting the threshold test from intent theory to scope theory. See supra __. 146 Collins, supra note __, at __. 147 If an algorithm were defined in a stand-‐alone manner and the functions 1, 2, and 3 remained an algorithm regardless of their context, the specification would have to disclose sub-‐algorithmic steps.



technology is not a harmless party trick. It reveals that section 112(f) operates in formalistic

manner when it curtails permissible patent scope. In an ideal patent regime that imposes

substantive limits on what can be patented, two claims that are coextensive—that is, two

claims that describe the same sets of technologies—should be subjected to the same

restrictions because they raise identical policy concerns. However, when operating on the basis

of a relative definition of structure, section 112(f) does not achieve this goal. The second claim

construed under the default rules of Phillips would be coextensive with the first claim

construed under scope-‐narrowing section 112(f).148 Yet, the relative definition of structure

means that the second claim, too, must be subject to section 112(f) and narrowed in scope to

the more granular, (sub-‐)algorithmic steps disclosed in the specification. The Federal Circuit’s

relative definition of an algorithm formalistically narrows the scope of every functional

software limitation, even if that limitation already recites a highly specific function.

B. The Pre-‐Williamson Threshold Test as a Model?

Before Williamson, the strong presumptions animated by intent theory did most of the

work in the section 112(f) threshold test in software cases.149 Nonetheless, the Federal Circuit

did occasionally entertain arguments that there was sufficient structure in a claim limitation

that recited “means” or insufficient structure in a claim limitation without the word “means,”

and these arguments were sometimes, although rarely, sufficient to carry the day. Because a

148 Actually, the literal scope of the second claim would be narrower because it does not include the equivalents of steps 1, 2, and 3. 149 Section 112(f) software cases only became pervasive after intent theory had become dominant, so the strong presumptions governed most of them.



relative definition of structure is useless in the section 112(f) threshold test,150 these cases must

have employed a stand-‐alone definition of structure of some kind. As such, they should, in

theory, provide fodder for the stand-‐alone definition of structure that is needed after

Williamson to administer a threshold test based on scope theory.

In actuality, however, the pre-‐Williamson cases addressing the threshold test offer little

guidance for how to proceed in post-‐Williamson cases. Some opinions use conclusory

reasoning, baldly asserting that what amounts to a programmed computer is actually sufficient

structure.151 Where there is more than conclusory reasoning, the reasoning is either misguided

or protean. Section IV.B.1 addresses a first line of cases that derives from the use of the term

“circuit” as a claim limitation and that unhelpfully reverts to the notion of software’s physical

structure. Section IV.B.2 turns to a second, and much more sparse, line of cases suggesting that

functional claim limitations describing an operation’s “input, output, and connections” or “how

its output may be achieved” should count as logical structure under the section 112(f) threshold

test. This definition of software’s structure is radically underdeveloped, but it at least hints at

what would be needed to formulate a stand-‐alone definition of structure.

1. Circuit as Physical Structure

A number of the pre-‐Williamson cases in which the Federal Circuit found sufficient

structure in a claim limitation that lacked the term “means” to overcome the presumption

against the application of section 112(f) derive from a single precedent: Apex Inc. v. Raritan 150 See supra Section IV.A.2. 151 LG Elecs., Inc. v. Bizcom Elecs., Inc., 453 F.3d 1364, 1372 (Fed. Cir. 2006), rev’d on other grounds, Quanta Computer, Inc. v. LG Elecs., Inc., 553 U.S. 617 (2008) (“The claim itself provides sufficient structure, namely ‘a CPU and a partitioned memory system,’ for performing the stated function, ‘controlling the communication unit.’”).



Comp.152 Apex involved a patent on “computerized switching systems, known in the art as

keyboard, video, mouse … switches, that allow centrally located network administrators to

operate multiple server computers without requiring a complicated wiring scheme.”153 The

claim contained, inter alia, a “programmable logic circuit” limitation for performing a variety of

functions, including “control[ling] the programmable switch in response to the keyboard or

cursor control device signals detected.”154 The district court held that this limitation, in addition

to a number of others that also used the term “circuit,” were governed by section 112(f).155 The

Federal circuit vacated the district court’s holding, stating that the presumption against section

112(f) had not been overcome.156 The Federal Circuit’s reasoning relied on the fact that a circuit

has physical structure because devices must be connected in a path:

The term “circuit” is defined as “the combination of a number of electrical devices and conductors that, when interconnected to form a conducting path, fulfill some desired function.” Dictionary of Computing, 75 (4th ed.1996). In light of this definition, it is clear that the term “circuit,” by itself connotes some structure. In the absence of any more compelling evidence of the understanding of one of ordinary skill in the art, the presumption that § 112, ¶ 6 does not apply is determinative.157

The Federal Circuit did “not find it necessary to hold that the term ‘circuit’ by itself always

connotes sufficient structure,” but rather held that “the term ‘circuit’ with an appropriate

152 Apex Inc. v. Raritan Comp., Inc., 325 F.3d 1364, 1373–74 (Fed. Cir. 2003). 153 Id. at 1367. 154 Id. at 1368. 155 Id. at 1369. 156 Id. at 1371-‐75. 157 Id. at 1373. For additional definitions of circuit that require an electrical pathway, see Linear Tech. Corp. v. Impala Linear Corp., 379 F.3d 1311, 1320 (Fed. Cir. 2014).



identifier such as ‘interface,’ ‘programming’ and ‘logic,’ certainly identifies some structural

meaning to one of ordinary skill in the art.”158

Based on its holding in Apex as precedent, the Federal Circuit proceeded in future cases

to find sufficient structure in a number of software limitations lacking the word “means.” Some

of these limitations used the term “circuit,” just as Apex did.159 Other limitations, however, did

not, and the Federal Circuit only analogized a software limitation to a hardwired electrical

circuit structured in a path.160 Whatever merit this invocation of the physical structure of a

circuit has in relatively simple hardware devices, it has little to no relevance for software

limitations in a post-‐Williamson world. The word “circuit” simply launders function into

structure.

The oddity of the Federal Circuit’s circuitry doctrine is on full display in the Federal

Circuit’s opinion in MIT v. Abacus Software.161 The patent at issue claimed a system for

producing color copies. Two of the claim’s limitations were “aesthetic correction circuitry for

interactively introducing aesthetically desired alterations into [the] appearance signals to

produce modified appearance signals,” and “colorant selection mechanism for receiving said

modified appearance signals and for selecting corresponding reproduction signals … to produce

158 Apex, 325 F.3d at 1373. 159 Linear Tech., 379 F.3d at 1320–21; MIT v. Abacus Software, 462 F.3d 1344, 1353–55 (Fed. Cir. 2006); Power Integrations, Inc. v. Fairchild Semiconductor Int’l, Inc., 711 F.3d 1348, 1365 (Fed. Cir. 2013). 160 Inventio AG v. ThyssenKrupp Elevator Americas Corp., 649 F.3d 1350, 1358 (Fed. Cir. 2011) (“In this case, the claims indicate that the ‘modernizing device’ functions as an electrical circuit that receives signals, processes signals, and outputs signals to other components in the patented system.”). 161 MIT v. Abacus Software, 462 F.3d 1344 (Fed. Cir. 2006).



… a colorimetrically-‐matched reproduction.”162 The Federal Circuit held that the former

limitation connoted sufficient structure for the presumption against section 112(f) not to be

overcome because “the term ‘circuitry,’ by itself, connotes sufficient structure” and that, again

laundering function into structure, the “aesthetic correction” language, although functional,

“adds further structure by describing the operation of the circuit.”163 In contrast, the

presumption against section 112(f) was overcome for the latter limitation because the

limitation failed to connote sufficiently definite structure. The Federal Circuit held that “[t]he

term ‘mechanism’ standing alone connotes no more structure than the term ‘means,’” and that

the “colorant selection” modifier did not have a generally understood meaning in the art that

connoted structure, either.164 Here, the circuitry doctrine is simply an example of “the doctrine

of magic words” in action.165 There is nothing that is meaningfully more structural—at least,

when structure is understood as physical structure—about the “aesthetic correction circuitry”

limitation than the “colorant selection mechanism” limitation.

2. Inputs and Outputs as a Clue to Logical Structure

A second theme in the Federal Circuit’s pre-‐Williamson cases addressing the section

112(f) threshold test offers a more promising starting point for developing a stand-‐alone

162 Id. at 1348. 163 Id. at 1355–56. The dissent argued that the limitation was a section 112(f) limitation. [] In response, the majority’s rejoinder is that “the dissent appears to misapprehend the strength of the presumption that applies when the term ‘means’ does not appear in the claim.” Id. at 1356. 164 Id. at 1353–55. 165 Julie E. Cohen & Mark A. Lemley, Patent Scope and Innovation in the Software Industry, 89 Cal. L. Rev. 1, 9 (2001) (describing a rule under which software was patentable so long as the applicant recited “magic words and pretended that she was patenting something else entirely”). The legal significance of the “circuit” term could be explained by intent theory, but that is not a rationale that the Federal Circuit has ever offered. Rather, the Federal Circuit has insisted that “circuit” connotes physical structure.



definition of software’s logical structure. This theme was first overtly articulated in Judge

Reyna’s majority opinion in Apple v. Motorola, issued only a year before Williamson.166 In the

course of explaining why the strong presumption against applying section 112(f) to a limitation

not using the word “means” was not rebutted, Judge Reyna asserted that limitation recited

sufficient structure because it described the limitation’s “input, output, and how its output may

be achieved.”167 In so doing, Judge Reyna laid some preliminary groundwork for a stand-‐alone

definition of logical structure in software.

Among the many patents and issues addressed, Apple upheld several apparatus claims

in Patent No. 7,479,949 on the use of finger contacts change the visual field or select files a

touchscreen computer.168 Each of the claims at issue describes one or more “heuristics” for

determining that certain finger contacts correspond with certain commands. For example, one

limitation recited “a vertical screen scrolling heuristic for determining that one or more finger

contacts correspond to a one-‐dimensional vertical screen scrolling command rather than a two-‐

dimensional screen translation command based on an angle of initial movement of finger

contact with respect to the touch screen display.”169

166 Apple Inc. v. Motorola Inc., 757 F.3d 1286 (2014). 167 Id. at 1300. 168 Id. at 1294–1304. The Apple opinion is much better known for its implications for injunctions for standard-‐essential patents. Id. at 1331–32. 169 Id. at 1295. Another limitation recites the inverse heuristic that works in tandem with the heuristic in the text: “a two-‐dimensional screen translation heuristic for determining that … one or more finger contacts correspond to a two-‐dimensional screen scrolling command rather than a one-‐dimensional vertical screen translation command based on an angle of initial movement of finger contact with respect to the touch screen display.” Id. A final limitation recites “a next item heuristic for determining that … one or more finger contacts correspond to a command to transition from displaying a respective item in a set of items to displaying a next item in the set of items.” Id.



The district court construed the term “heuristics” to mean “one or more rules to be

applied to data to assist in drawing inferences from that data.”170 Applying the section 112(f)

threshold test, the district court then held the heuristics limitations to be subject to section

112(f), despite the then-‐present strong presumption that a limitation not using the world

“means” is not a means-‐plus-‐function limitation, because the claims described functions

“without describing the structure necessary to perform the functions.”171 According to the

district court, the limitation would encompass any rule for distinguishing touches that generate

vertical, one-‐dimensional screen scrolling from touches that would generate diagonal, two-‐

dimensional screen scrolling.172 Finally, the district court identified the corresponding structure:

“a heuristic that uses as one input the initial angle of the user’s finger swipe gesture and

170 Id. Judge Posner sat by designation as the district court judge. Id. at 1294. 171 Id. at 1295. At least in part, Judge Posner’s reasoning in the threshold test seems to employ the very category error noted above that can be implicated by using the Federal Circuit’s definition of an algorithm as a multi-‐step process to identify sufficient structure in a singular claim limitation. See supra __. When assessing whether each functional heuristic limitation in the claim individually was subject to section 112(f), Judge Posner identified an algorithm as the structure that must exist in a claim limitation to avoid section 112(f). Apple Inc. v. Motorola, Inc., No. 1:11-‐cv-‐08540 at 14–15 (N.D. Ill. March 19, 2012) (positioning an “algorithm, or ‘step-‐by-‐step process,’ for performing [the claimed] function” as “the ‘structure’ required by statute” in the claim to avoid the strictures of section 112(f)). Judge Posner concluded that “[t]he claims in patent ‘949 gesture toward such a step-‐by-‐step process, but don’t describe one.” It is not clear what it means for a single function recited in a claim limitation to be a multi-‐step algorithm (let alone what it means to “gesture toward” one). Apple Inc. v. Motorola, Inc., No. 1:11-‐cv-‐08540 at 7 (N.D. Ill. Jan. 16, 2012). 172 In its reasoning explaining why the limitation is purely functional and lacks sufficient structure, the district court technically over-‐states the breadth of the limitation:

It is inherent in the concept of a touch screen computer that the user would want to enter commands to tell it to do things, that he would enter those commands by moving his fingers on the touch screen, and that the device would need to apply some set of rules to a given finger input to determine which command was intended. Apple’s patent cannot cover every means of performing the function of translating user finger movements into common computer commands on a touch-‐screen device—that would be a patent on all touch-‐screen computers.

Apple Inc. v. Motorola, Inc., No. 1:11-‐cv-‐08540 at 14–15 (N.D. Ill. March 19, 2012). The claim does not “cover every means of performing the function of translating user finger movements into common computer commands on a touch-‐screen device.” Id.



determines whether that angle is within a predetermined range of being perfectly vertical, as

shown for example in Figure 39C.”173

In the Federal Circuit’s majority opinion, Judge Reyna reversed the district court, holding

that the claim limitation was not a section 112(f) limitation.174 Judge Reyna concluded that the

limitation connoted structure because it recited “the heuristics’ operation within the context of

the invention, including the inputs, outputs, and how certain outputs are achieved.”175 This

inputs-‐and-‐outputs line of reasoning is problematic in some ways. Its support in Federal Circuit

precedent is weak.176 Furthermore, the claim of the ’949 patent at issue in Apple arguably did

not include inputs and outputs as limitations on claim scope. Judge Reyna arguably relied

excessively on features of the preferred embodiments disclosed in the specification that do not

actually limit claim scope to find that the claim language connoted structure.177 Nonetheless,

173 Apple Inc. v. Motorola, Inc., No. 1:11-‐cv-‐08540 at 7 (N.D. Ill. Mar. 29, 2012). 174 Apple, 757 F.3d at 1294–1304. 175 Id. at 1302; see also id. at 1299 (“Structure may … be provided by describing the claim’s operation, such as its input, output, or connections.”); id. at 1300 (“We find … that the ’949 patent … describes the limitation’s operation, including its input, output, and how its input may be achieved.”); id. at 1303 (The limitation “describes [the invention’s] operation, including its input, output, and how its output may be achieved.”). Judge Reyna also argued “that ‘heuristic’ has a known meaning” that connotes structure. Id. at 1300. The opinion eventually concludes that “[w]e need not decide here whether the term ‘heuristic,’ by itself, connotes sufficient structure to maintain the presumption against means-‐plus-‐function claiming” because the inputs-‐and-‐outputs argument was sufficient to demonstrate that the strong presumption against section 112(f) had not been rebutted. Id. at 1301. 176 Apple cites two cases to support its looking to the “input, output, and connections” of a program as a clue for identifying structure in a claim limitation. Id. at 1299. First, Linear Tech. Corp. v. Impala Linear Corp., 379 F.3d 1311, 1320–21 (Fed. Cir. 2014), held that the presumption against section 112(f) was not overcome principally because the term “circuit” connoted tangible structure. See supra __ (discussing the section 112(f) “circuit” cases). However, Linear did note in passing that the term “circuit” became more structural when coupled in the claim with a description of the circuit’s “operation” and its “desired output.” Linear, 379 F.3d at 1310. Second, Lighting World, Inc. v. Birchwood Lighting, Inc., 382 F.3d 1354, 1361–62 (Fed. Cir. 2004), held that the presumption against section 112(f) was not overcome principally because the term “connector” had a generally understood meaning in the art as a name for physical structure. 177 The claim language clearly recited the heuristic’s “objectives.” Id. at 1302. Beyond that, however, all of the opinion’s support for the inputs, outputs, and connections as claim limitations drew from the written description. Id. at 1302–03. One of the dissent’s critiques of the majority is that the majority relies too heavily on the preferred



despite these shortcomings, Judge Reyna’s opinion is interesting because, unlike the rest of the

Federal Circuit’s pre-‐Williamson jurisprudence on the threshold test, it suggests one factor that

might be relevant in developing a stand-‐alone definition of software’s structure.178

The gist of Judge Reyna’s reasoning seems to reside in the notion that functional claim

language should count as a description of the logical structure of software if the function

recited is sufficiently specific that describes how the invention works—that is, which it

describes “inputs, outputs, and how certain outputs are achieved”—rather than what the

invention does for its user.179 This approach to answering the threshold test’s definitional

question is underdeveloped, but it at least hints at what is needed to move forward after

Williamson, and its shortcomings are arguably endemic to any answer to the definitional

question that formulates a stand-‐alone, rather than relative, definition of structure in software

limitations.

embodiments in the specification in the section 112(f) threshold test. Id. at 1335 (Prost, J., concurring-‐in-‐part and dissenting-‐in-‐part) (“In effect, what the majority has done is imported the second step of the analysis (where you define the scope of a means-‐plus-‐function claim based on the corresponding structure in the specification) into the first step (where you identify whether the term is drafted in means-‐plus-‐function format.”). This type of disagreement should not be surprising given that the Federal Circuit has not been clear about whether Phillips governs the search for sufficient structure in section 112(f) threshold test. See supra __. 178 At least one district court has used Apple’s inputs-‐and-‐outputs reasoning to find sufficient structure in a claim limitation lacking the word “means” to avoid section 112(f) status. Finjan, Inc. v. Proofpoint, Inc., 2015 WL 7770208, at *11 (N.D. Cal. Dec. 3, 2015) (“The term ‘content processor’ has a sufficiently specific structure. Independent Claim 1 describes how the ‘content processor’ interacts with the invention’s other components (the transmitter and receiver), which informs the term’s structural character…. [T]he intrinsic evidence establishes the structural character of ‘content processor’ through its interaction with the system’s other components.”). 179 The language used to convey this reasoning is a bit awkward. As the opinion puts it, “[s]tructure may also be provided by describing the claim’s operation, such as its input, output, or connections. The limitation’s operation is more than just its function; it is how the function is achieved in the context of the invention.” Apple, 757 F.3d at 1299. This language is misleading because it suppresses the fact that a description of “how the function is achieved in the context of the invention” is itself inevitably a description of function. See supra __. A more accurate way of describing the needed line in a stand-‐alone definition of structure in software is that the logical function that serves as structure is function, too, but that it is just a type of function that is sufficiently specific that it should count as metaphorical structure.



V. Towards a Stand-‐Alone Definition of Software’s Structure

Williamson mandates a revolution in the law of section 112(f) as applied to software,

but it is as-‐of-‐yet unclear which of two possible revolutions it will yield.180 First, if the Federal

Circuit were to import its already developed, relative definition of the structure of a software

invention from the search corresponding structure in the specification into the threshold test,

then the outcome, but not the doctrine, would be revolutionary. All functional software

limitations would be subject to section 112(f), regardless of whether they employ the term

“means.”181 Second, if the Federal Circuit were to develop a stand-‐alone definition of structure

for the threshold test, then it would need a revolution in its section 112(f) doctrine. To date, the

Federal Circuit has not sought to develop the stand-‐alone definition of structure that is needed

for the threshold test to draw a principled line that subjects some, but not all, functional

software limitations to section 112(f), although the discussion of inputs, outputs and

connections in Apple v. Motorola hints at factors that could be considered as part of such a

definition.182 The remainder of this essay looks down the path that the Federal Circuit would

have to follow if it were to meet Williamson’s challenge by revolutionizing its section 112(f)

doctrine and developing a stand-‐alone definition of software’s structure for the threshold test.

In Williamson, the Federal Circuit recognized that the policy concerns about

overbreadth expressed in the Supreme Court’s functional-‐claiming cases from the first half of

180 These two options ignore the misguided possibility that physical structure is a relevant consideration outside of exceptional cases. See supra __. 181 See supra __. 182 See supra __.



the twentieth century require a threshold test motivated by scope theory.183 When there is

enough structure recited in the limitations of a claim, the number of “other devices beyond our

present information or indeed our imagination which will perform [the claimed] function and

yet fit [the] claims” is sufficiently reduced to ward off the most serious social costs of overbroad

patents.184 The same policy concerns should frame the answer to the sui generis definitional

“What is structure?” question of the threshold test in software: At what level of claim

specificity are the restraints on competition engendered by functional software claims

acceptable because the claim is restricted in scope to particular means for achieving a useful

end?

The basic premise of a stand-‐alone definition of structure in software must be a line on

a ladder of generality or abstraction. Functional descriptions of a software invention exist at

many different levels of generality/abstraction, ranging from extremely specific (e.g., an in-‐

depth description of modules and their interaction) to extremely general (e.g., a description of

the utility enjoyed by the end-‐user of the software). A stand-‐alone definition of structure must

identify the point on this ladder at which a granular, functional description of a software

program that constitutes logical structure transitions into a highly general, structureless

description of a software program.185 The descriptions both above and below the line are

technically functional descriptions in a linguistic sense, but descriptions at or below the line

183 See supra __. 184 Halliburton Oil Well Cementing Co. v. Walker, 329 U.S. 1, 12 (1946). 185 In a sense, the term “stand-‐alone structure” is misleading. The identification of structure still requires a comparison to a baseline: Is the functional description above or below the point at which the transition occurs? However, this baseline is fixed. Unlike a comparison to the baseline provided by a claim, the baseline does not shift dramatically from one section 112(f) analysis to the next.



count as metaphorical structure for the purposes of section 112(f). To reiterate, this line on the

ladder of functional abstractions is not needed in the mechanical arts because section 112(f)

can put all functional descriptions in the same bucket and require a dollop of physical structure

as a limitation on permissible claim scope. However, because the physical, structural properties

of a software invention are not relevant when describing what a software inventor has

accomplished, the ladder-‐of-‐abstraction approach that discriminates among functional

descriptions is the only way of developing the stand-‐alone definition of logical structure that is

needed if section 112(f) is to subject some, but not all, software limitations to its scope-‐

narrowing rule of claim construction.

In his work on functional claiming in software, Mark Lemley argues that section 112(f)

must make a distinction between a “goal” (not structure) and a “way of implementing a goal”

(structure),186 or a “problem” the patentee solved (not structure) and “what the patentee …

actually did” to solve the problem (structure).187 The difficulty with this distinction is, of course,

that goals and problems can be expressed in functional language couched at many different

rungs on the ladder of abstraction. Your goal is likely to be my way of implementing a different

goal. A software inventor may be motivated by the problem that two modules interact in an

inefficient manner or by the problem that the parties to a transaction face settlement risk.188

186 Lemley, supra note __, at 947 187 Id. at 963. 188 Cf. Alice Corp. v. CLS Bank Int’l, 134 S.Ct. 2347 (2014) (holding that a claim to a computer-‐executed method of reducing settlement risk is patent-‐ineligible). The interaction of the Federal Circuit’s Williamson opinion and the Supreme Court’s Alice opinion raises interesting questions that are beyond the scope of this essay. Nonetheless, it is worth noting that a stand-‐alone definition of software’s structure developed to implement Williamson might shed some light on when a patent describes a claim “purport[s] to improve the functioning of the computer itself” and thus is patent-‐eligible even if it is directed to an abstract idea. Id. at 2359. Perhaps improvements in logical structure are examples of improvements in the functioning of the computer itself.



What is needed in the post-‐Williamson world is a stand-‐alone definition of what constitutes a

structureless problem or goal and, inversely, what constitutes a way of implementing a goal or

a patentee’s actual solution to a problem that should count as logical structure.

Importantly, once formulated, a stand-‐alone definition of structure reaches the same

conclusion about whether a functional description counts as logical structure regardless of

whether the description is a claim limitation (and the question is whether the claim connotes

sufficient structure to avoid section 112(f)) or text in a specification (and the question is

whether the specification discloses corresponding structure). The exact same definition of

structure could apply in both contexts.189 Furthermore, when used to identify corresponding

structure, it should not make the existence of structure contingent on the claim limitation at

issue. Some specifications that disclose corresponding structure under the Federal Circuit’s

definition of an algorithm will fail to disclose corresponding structure under the new stand-‐

alone definition of software’s structure. Even though a functional description in the

specification is more granular than the functional description recited in the claims, the

specification’s functional description may still be too high up on the ladder of generality.

Conversely, some specifications that fail to disclose corresponding structure under the Federal

Circuit’s definition of an algorithm because they simply repeat the functions that were recited

as claim limitations will disclose structure under the stand-‐alone definition of software’s

structure. The functional description recited in both the claims and the specification may be

sufficiently far down the ladder of generality to count as logical structure. In these cases,

189 In theory, the Federal Circuit could adopt a stand-‐alone definition of structure for the threshold test and continue to use its relative definition of structure to identify corresponding structure in the specification. However, once the stand-‐alone definition has been formulated, it makes little sense not to use it in both doctrinal contexts.



however, the presence of corresponding structure in the specification will technically be

irrelevant because the presence of structure in the claims means that the limitations will not be

subject to section 112(f) in the first place.

Drawing the line that is needed to identify logical structure in the software arts will not

be an easy task. The echoes of Learned Hand’s levels-‐of-‐generality test for drawing the

idea/expression dichotomy in copyright are clear, and that test is notorious for its lack of ex-‐

ante clarity.190 In fact, it is the difficulty of a parallel task that led the Federal Circuit to rely on a

threshold test motivated by intent theory to shelve the project of bringing section 112(f) to

bear on step-‐plus-‐function limitations in method claims, despite the statutory requirement to

do so.191 In method claims, there is no clear, intuitive difference between a “step” and an “act,”

just like there is no clear, intuitive difference between functionally defined software limitations

that should, and should not, be subject to section 112(f).192 The only difference is a functional

description’s position on a ladder of generality. An “act” describes a more specific functional

task than a “step” does, and several acts must be strung together to explain how to perform a

step. When is a functional operation recited as a claim limitation sufficiently specific to qualify

as an act rather than a step? The Federal Circuit punted on this question because of its

difficulty, leaning heavily on intent theory to create an even stronger than usual presumption in

190 Nichols v. Universal Pictures Corp., 45 F.2d 119 (2d Cir. 1930). The actual copyright doctrine that has developed in software copyright cases is helpful. In copyright, the idea/expression dichotomy seeks, among other goals, to prevent copyright from granting an author control over software functionality at any level of abstraction. Computer Associates Int’l v. Altai, Inc., 982 F.2d 693 (2d Cir. 1992). In patent law, section 112(f) must draw a line between claims to functionality that are sufficiently specific and those that are excessively general. 191 See supra notes __ and accompanying text. 192 Judge Rader recognized that the distinction between a step and an act is “inherently more problematic” than the distinction between a function and a structure (at least in non-‐software claims. Seal-‐Flex, Inc. v. Athletic Track and Court Const., 172 F.3d 836, 848–49 (Fed. Cir. 1999) (Rader, J, concurring).



the section 112(f) threshold test that method claim limitations without the (rarely used) term

“step” are not step-‐plus-‐function limitations.193 Yet, this question parallels precisely the

question that Williamson has now teed up for the Federal Circuit to answer in order to create a

stand-‐alone definition of software’s structure.194

The line marking a binary legal distinction on a continuous spectrum has no “natural”

position on the ladder of abstractions. Nor is it feasible to ask the ultimate economic

question—At what level of abstraction does a patent claim provide sufficient, but not excessive,

reward to a software innovator?—as part of each and every patent examination procedure and

invalidity defense. What is needed is an administrable proxy to replace the proxy of physical

structure that is used in mechanical technologies but that is not available in software. The

author has suggested elsewhere that one possible proxy uses consumer preferences as an

anchor.195 Perhaps descriptions of functionality that map onto the reasonable consumer’s

desires should be labeled as goals because these preferences contribute to the definition of

product markets, and perhaps descriptions of functionality that map onto particular ways of

satisfying those desires should be seen as ways of achieving those goals and thus as logical

structure. This proxy is far from straight-‐forward, but it at least provides an objective reference

to which the stand-‐alone definition can be pegged.

193 Id. at 849 (“The difficulty of distinguishing acts from functions … places a significant burden on the claim drafter to choose language with a definite and clear meaning. To invoke a presumption of § 112, ¶ 6 application, a claim drafter must use language that expressly signals the recitation of a function as distinguished from an act.”). 194 In fact, if Williamson applies to both apparatus and method claims, the Federal Circuit can no longer rely on a strong presumption against the application of section 112(f) to avoid having to draw a distinction between functions and acts in method claims. 195 Kevin Emerson Collins, Patent Law’s Functionality Malfunction and the Problem of Overbroad, Functional Software Patents, 90 Wash U. L. Rev. 1399, 1421–23, 1466 (2013).



The next step required to develop a stand-‐alone definition of structure for software

under section 112(f) involves an interdisciplinary conversation among patent lawyers, computer

scientists, and economists.196 The desired output of such a conversation is a menu of levels of

generality at which software can be described that is grounded in computer science, and an

analysis of the implications of selecting any one of these levels—or a combination of them in

different contexts—as a stand-‐alone definition of logical structure in software for the purpose

of section 112(f). With a menu of options identified, and the consequences of choosing the

different options clarified, the Federal Circuit will be well positioned to undertake the

revolution in the section 112(f) law of software patents that Williamson mandates.

To be clear, the proposed conversation may not flush out any possible distinctions

grounded in computer science that provide an acceptable answer to the definitional “What is

structure?” question. It may be that, at the end of the day, the lesson learned from the

conversation is that there are no viable distinctions upon which to build a reformed law of

software patents that produce anything close to the desired amount of clarity. The purely

functional nature of software may mean that section 112(f) just cannot regulate software

patents in the way that it regulates mechanical patents.197

196 Collins, supra note __, at __. Pam Samuelson’s efforts at revamping intellectual property protection for computer software provide an interesting model of the needed conversation, but not of the desired outcome because the scope of the task at hand after Williamson is quite different from the scope of the task that Samuelson undertook. Pamela Samuelson et al., A Manifesto Concerning the Legal Protection of Computer Programs, 94 COLUM. L. REV. 2308, 2323 (1994). 197 Cf. Kevin Emerson Collins, Patent-‐Ineligibility as Counteraction, Wash U. L. Rev. (forthcoming 2017) (arguing that purely functional technologies cause regulatory inefficacy in doctrines that, like section 112(f) and written description, rely on the recitation of structure to limit the overbreadth of functional claims).



Conclusion

Williamson v. Citrix Online altered the section 112(f) threshold test for determining

whether a functional claim limitation that not use the term “means” should be construed using

the scope-‐narrowing rules of means-‐plus-‐function claiming. More specifically, Williamson

expands the reach of section 112(f) by replacing a formalistic, strong-‐presumption test with a

substantive, sufficient-‐structure test.

On its face, Williamson may only appear to implicate the quantitative “How much

structure is enough?” question of the threshold test. That is, Williamson may only appear to

require a bit more structure in the claim limitations than was previously required in order to

evade section 112(f). However, with respect to software patents in particular, Williamson

actually signals a revolution because it will finally force the Federal Circuit to address the

definitional “What is structure?” question in software.

One understandable, initial reaction to Williamson’s impact on software in the patent

community might be a collective shrug of the shoulders. Before Williamson, there was already a

well-‐established answer to the definitional question of software’s structure in the context of

identifying corresponding structure in the specification: software’s structure is an algorithm, or

step-‐by-‐step procedure, for performing the function recited as a claim limitation. However, this

definition of an algorithm is relational, and a relational definition of structure is of little use in

the Williamson threshold test. A reader cannot examine a series of functions in isolation and

say that it is or is not an algorithm. Whether a specification recites an algorithm can only be

determined by comparing the series of functions disclosed in the specification to the baseline



functionality recited as a claim limitation. Corresponding structure is defined only in relation to

a particular claim; the exact same disclosure may be structure in relation to a more general

claim and not structure in relation to a more specific claim. As a definition of corresponding

structure, this relative definition of structure in software provides workable rule. However,

when searching for structure in the claims, it is highly problematic because there is no extant

baseline to which to compare the functional limitations.

Thus, Williamson will require a significant shift from the pre-‐Williamson conceptions of

what constitutes the structure of a software invention. This shift may take either one of two

different forms. Recognizing that an algorithm only provides a metric for identifying software’s

structure in the specification, the Federal Circuit could decide that software claims themselves

never connote structure and that all functional software limitations are thus subject to section

112(f). Alternatively, the Federal Circuit could undertake the difficult task of articulating a

novel, stand-‐alone definition of software’s structure by drawing a line on a ladder of

abstractions much like the idea/expression dichotomy does in copyright law. Either way,

however, Williamson marks a revolution in the law of section 112(f) as applied to software.

Documents

Kevin Emerson Collins, The Williamson Revolution in Software's